WO2000016259A1 - Dispositif visuel - Google Patents
Dispositif visuel Download PDFInfo
- Publication number
- WO2000016259A1 WO2000016259A1 PCT/JP1999/004975 JP9904975W WO0016259A1 WO 2000016259 A1 WO2000016259 A1 WO 2000016259A1 JP 9904975 W JP9904975 W JP 9904975W WO 0016259 A1 WO0016259 A1 WO 0016259A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- image
- band pixel
- pixel value
- edge information
- moving
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
- G06V20/695—Preprocessing, e.g. image segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/12—Edge-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10056—Microscopic image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30024—Cell structures in vitro; Tissue sections in vitro
Definitions
- the present invention relates to a visual device and a method for counting the number of objects in an image, and more particularly, to three primary color wavelengths, visible light wavelengths, infrared wavelengths, ultraviolet wavelengths, and all other electromagnetic waves captured by a video camera or a digital camera.
- the present invention relates to an apparatus for counting the number of moving or stationary objects in an image composed of arbitrary bands. Background art
- a typical example is a device for counting the number of cells taken through a microscope. Since the chromosomes of cells are stained purple, the number of cells can be counted by cutting out a purple region of a certain size or more from the image into one clump. However, when it is difficult or impossible to stain cells, it is not easy to count the number of cells. Because cells are generally transparent, color information is not very useful for making the whole cell into one clump. Of course, magnifying a cell image can capture nuclei and mitochondrial shadows, but this is the exception.
- edge information is often generated from the contours of cells projected by refraction and reflection of light. If this edge information is used, the whole cell could theoretically be made into one lump. However, in most cases, the edge information is incomplete in most cases, so the edge information is obtained using information such as the shape and size of the cell. Is complementary. Moreover, in order to make the whole cell into one clump from the edge information, it is necessary to perform image processing with a large amount of calculation, such as filling, and it is natural that the edge information must not have a break. In addition, if only cells that are moving from cells are selected and cut out, calculations such as optical flow must be performed.In order to improve counting accuracy, expensive equipment is required, while inexpensive equipment is required. If you use a different device, the calculation time becomes enormous.
- the visual device recognizes the change in color information as the movement of the object, the performance of the illumination and force measurer will not be a problem, but it will be difficult to accurately reproduce the shape of the object from the movement of the object, but it will be surrounded by force edge information.
- the problem of having to determine the object area by filling in the shaded area has arisen, and visual devices utilizing changes in color information have not been sufficiently studied.
- the visual device can capture the movement of the object from changes in color information and generate edge information and cut out the object from this edge information into one lump, And the number of objects can be counted without depending on the shooting environment. If the edge information can be generated from the color information by vibrating the object itself, vibrating the camera, or otherwise vibrating the captured image, the visual device can be used even if the object is stationary It is expected that the number of objects can be counted.
- the visual device capable of counting the number of objects photographed by a camera. If the object photographed by the camera is stationary, the visual device can always count the number of stationary objects. However, if the object is moving, the visual device can only count the number of moving objects while the camera is shooting the moving object. There is no problem if the positions of stationary and moving objects are specified in advance, such as cells in a petri dish. Space is a camera image Fixing the camera limits the use of the vision device because the camera does not fit into the corner or humans and animals grow and shrink in the image depending on the distance from the camera. In addition, since the visual device must distinguish between the interior of the room and humans or animals, it requires a large amount of computation to recognize objects.
- Mobility Mera searches for specific objects, such as humans and animals, from inside or outside the room, and photographs only those objects, so that these objects become the appropriate size in the image. If the visual device can adjust the magnification of the camera, the visual device can not only recognize these objects easily, but also detect moving objects such as humans and animals whose positions cannot be specified in advance. You will be able to count numbers. Of course, even if humans and animals are sleeping and hardly move, visual devices are expected to be able to count humans and animals by distinguishing humans and animals from other stationary objects.
- the present invention described in the appended claims provides a method for counting the number of moving objects or all objects at high speed based on edge information generated by selectively selecting either moving objects or all objects in a moving image.
- the purpose is.
- Another object is to count the number of moving and stationary objects from a moving image to calculate the ratio of moving and stationary objects in the moving image at high speed.
- it aims to count the number of moving and stationary objects at high speed by searching for moving and stationary objects that exist within the range that can be photographed by the moving camera.
- a means for acquiring a frame image of the moving image for a moving object in the moving image a means for sequentially storing the frame image as a digital image, and a moving object from the digital image
- Means for generating a rough edge information image means for generating a moving object formation edge information image from the moving object rough edge information image using the digital image; and movement separated by the moving object body formation edge information image.
- the means for acquiring the frame image in a case where the moving image is an analog signal, the frame image is converted into a digital signal to be a digital image.
- the means for generating the moving object rough edge information image from the digital image includes: The edge information of the moving object is generated in the moving object rough edge information image using the digital image.
- the means for generating a body-forming edge information image uses the digital image to form the edge information of the moving object into more accurate and clear edge information.
- the means for counting the number of moving object areas counts the number of the pixels representing the moving object area
- the means for holding the number of moving object areas indicates the moving object area
- the number of pixels to be output is output in a format required by the output destination, such as two's complement representation or floating point representation.
- the means for generating the moving object rough edge information image, the means for generating the moving object formation edge information image, and the means for detecting the position and the size of the moving object area are each:
- the visual device can be operated in parallel on a pixel-by-pixel basis, and all of these means are realized by local processing, so that the visual device does not depend on the shape or size of the moving object.
- the edge information is less affected by lighting and the like, and the movement is transparent, as compared with the case where the moving object area is extracted by classification of color information or the like. Since the edge information is generated by using refraction and reflection without dyeing even the object, the application range of the visual device is wide. Therefore, the problems relating to the counting of the moving objects are suitably solved.
- the invention according to claim 2 is a method for acquiring a frame image of a moving image for a moving object in the moving image, means for sequentially storing the frame image as a digital image, and moving from the digital image.
- a visual device comprising: means for separating an object area; means for detecting the position and size of the moving object area; means for counting the number of moving object areas; and means for holding the number of moving object areas. .
- the present invention is obtained by adding the means for separating the moving object region from the background using the moving object forming edge information image to the invention according to claim 1.
- the edge information of the moving object is used as a boundary, and the pixels included in the background and the moving object region included in the moving object region are used. Pixels can be classified into different groups. Accordingly, in the means for detecting the position and the size of the moving object region, the center of gravity of the moving object region and the number of pixels included in the moving object region are represented by one pixel. .
- the means may be operated in parallel, respectively; the means for generating the moving object rough edge information image; the means for generating the moving object forming edge information image; and the means for separating the moving object area from the background.
- the means, and the means for detecting the position and the size of the moving object area can be operated in parallel inside each of the pixels. Moreover, all of these means are realized by local processing. Therefore, the visual device can quickly count the number of the moving object regions without depending on the shape and size of the moving object. Further, compared with a case where the moving object region is extracted by classification of color information or the like, the moving object region separated from the background by the edge information is hardly affected by lighting or the like, and furthermore, the moving object region is transparent to the moving object region. However, since the wedge information is generated by using refraction and reflection without staining, the applicable range of the visual device is wide. Therefore, the problems relating to the counting of the moving objects are suitably solved.
- the visual device by providing a means for vibrating the digital image, the number of all object regions is calculated for all objects instead of the moving objects.
- This is a visual device that is characterized by When the means for vibrating the digital image is added to the visual device according to claim 1, the following is obtained.
- the means for acquiring the frame image when the moving image is an analog signal, the frame image is converted into a digital signal to be a digital image.
- the moving image is the digital signal, if the image is compressed, the image is expanded. If the image is not compressed, the image is input as it is. Thereafter, the frame image is cut out from the moving image to be the digital image.
- the means for sequentially storing the frame image as the digital image stores all the pixels of the frame image in a memory while maintaining a two-dimensional phase relationship.
- the means for generating the all-object rough edge information image from the vibration image generates the vibration image by vibrating the digital image vertically or horizontally in image units or pixel units. Thus, the entire object in the vibration image looks as if it is moving.
- the means for generating the all-object rough edge information image from the vibration image generates the edge information of the all object for each pixel of the vibration image. In the means for generating the all-object forming edge information image from the all-object rough edge information image using the digital image, the edge information of the all objects is more accurately and clearly displayed using the digital image. The edge information is formed.
- the means for detecting the position and the size of the entire object region divided by the all object forming edge information image includes: the center of gravity position of the edge information at the boundary of the entire object region; The number of the edge information at the boundary of the object area is represented by one pixel.
- the means for counting the number of all object regions counts the number of the pixels representing the entire object region.
- the means for holding the number of all object areas, and the number of pixels representing the all object areas are output in a format required by an output destination such as a two's complement representation or a floating point representation.
- the means can be operated in parallel with each other, and the means for generating the all-object rough edge information image, the means for generating the all-object forming edge information image, and the position and the position of the all-object area
- the means for detecting the size can be operated in parallel on a pixel-by-pixel basis in each of the sections. Moreover, all of these means are realized by local processing. Therefore, the visual device can count the number of all object areas at high speed without depending on the shape and size of the all objects. Also, depending on the classification of color information, etc. Compared to the case of extracting the entire object region, the edge information is less affected by lighting and the like.- Even if the transparent object is not dyed, the edge information uses refraction or reflection. Therefore, the visual device has a wide application range. Therefore, the problems relating to the counting of all the objects are preferably solved.
- the means for vibrating the digital image is added to the visual device according to claim 2, the following is obtained.
- the edge information of the entire object is used as a boundary, and the pixel included in the background and the entire object region are included in the pixel.
- the means can be operated in parallel, respectively; the means for generating the all-object rough edge information image; the means for generating the all-object forming edge information image; and the position and the position of the all-object area.
- the means for detecting the size can be operated in parallel on a pixel-by-pixel basis in each of the sections. Moreover, all of these means are realized by local processing. Therefore, the visual device can quickly count the number of all object regions without depending on the shape and size of the entire object. Further, as compared with the case where the entire object region is extracted by color information classification, the entire object region separated from the background by the edge information is hardly affected by illumination or the like, and furthermore, the entire object region is transparent. Since the edge information is generated by using refraction and reflection without staining, the range of application of the visual device is wide. Therefore, the problems relating to the counting of all objects can be suitably solved.
- the present invention relates to claim 1. 4.
- the means for generating the moving object forming edge information image from the object rough edge information image, and the means for generating the all object forming edge information image from the all object rough edge information image using the digital image are shared.
- the means for detecting the position and the size can be shared, and the means for counting the number of moving object areas, and the means for counting the total number of object areas can be shared.
- the means for retaining the number of moving object areas and the means for retaining the total number of object areas can be shared. If there is a restriction on the amount of hardware, these means may be shared. Of course, if they do not share, the counting time can be shortened accordingly.
- the edge information is hardly affected by illumination and the like, and is transparent to the moving object and the entire object. Since the edge information is generated by using refraction and reflection without staining, the application range of the visual device is wide. Therefore, the problems relating to the counting of the moving object and the total object are suitably solved.
- the means for detecting the position and said size of the area can be shared, and the means for counting the number of moving object areas, and the means for counting the total number of object areas can be shared.
- the means for holding the number of moving object areas and the means for holding the total number of object areas can be shared. If there is a restriction on the amount of hardware, these means may be shared. Of course, if not shared, the counting time can be shortened accordingly.
- the moving object area and the entire object area separated from the background by the edge information are more affected by illumination and the like than in the case where the moving object area and the entire object area are extracted by classification of color information or the like. Since the edge information is generated by using refraction and reflection without being dyed even for the difficult and transparent moving object and the entire object, the application range of the visual device is wide. Therefore, various problems relating to counting of the moving object and the total object are suitably solved.
- a function of retaining the number of stationary object areas obtained by subtracting the number of moving object areas from the number of all object areas, or the ratio of the moving objects, and the ratio of the moving objects It can have a function to keep the ratio and
- the number of stationary object regions can be obtained by subtracting the number of moving object regions from the total number of object regions.
- the means for holding the number of all object regions can output the number of stationary object regions instead of the number of all object regions as needed.
- the ratio of the moving object can be obtained by dividing the number of moving object regions by the total number of object regions. Therefore, the means for holding the number of moving object regions can output the ratio of the moving object regions instead of the number of moving object regions as necessary.
- the number of moving object regions is calculated from the number of all object regions. Since the number of the stationary object regions can be obtained by subtracting, the ratio of the stationary objects can be obtained by dividing the number of the stationary object regions by the number of the total objects divided by the number of regions. Therefore, the means for holding the total number of object regions can output the ratio of the stationary object region instead of the total number of object regions as necessary.
- the moving image captured by the video camera is configured such that the moving object in the frame image is moving by connecting a plurality of the frame images. It is pretending to be. Accordingly, by continuing the still images created at different times or at different places by a digital camera, a scanner, or the like, it is possible to make the moving object in the still image appear to be moving. By using the digital force camera and the scanner in place of the video camera, it is possible to easily adjust the interval of the shooting time of the still image when counting the number of moving object regions of the moving object having a slow moving speed. Can be.
- the digital camera having a low resolution can be used instead of the video camera.
- An inexpensive and highly accurate visual device can be realized. Therefore, various problems relating to counting of the moving object and the total object are suitably solved.
- the invention according to claim 4 is means for acquiring a frame image of the moving image for an arbitrary object in the moving image photographed by the moving power camera, and sequentially storing the frame image as a digital image.
- Means a means for generating an arbitrary object rough edge information image from the digital image, a means for detecting the position and size of an arbitrary object area partitioned by the arbitrary object rough edge information image, Means for converting a magnification into a position in environmental coordinates; means for converting the position and the size of the arbitrary object area into the position in the environmental coordinates; and a means for converting the environmental coordinates with respect to a plurality of the arbitrary object areas.
- a visual device comprising a means for controlling the position of the environmental coordinate to be moved, means for generating a camera command controlling the transfer power camera, a.
- the moving force The visual device adjusts the orientation and the magnification of the moving lens so that the arbitrary object photographed by the lens is photographed at an appropriate size.
- a series of processes for generating edge information of the arbitrary object from the moving image and degenerating the edge information are all realized by local processing. This allows the visual device to quickly shoot the arbitrary object with an appropriate size without depending on the shape and size of the arbitrary object.
- the edge information is less affected by illumination or the like, so that the visual device has a wide application range. Therefore, the problems relating to the photographing of the arbitrary object are suitably solved.
- the invention according to claim 5 is the visual device according to claim 4, wherein the arbitrary object forming edge information image is generated from the arbitrary object rough edge information image using the digital image, and the arbitrary object forming edge is formed.
- Means for separating the arbitrary object region from the background using an information image means for normalizing the arbitrary object region, means for holding an arbitrary object normalized image, and means for recognizing the arbitrary object normalized image
- Means for holding a recognition result means for generating an environment map represented by the environment coordinates; means for holding the environment map; means for estimating the position of the arbitrary object on the environment map;
- a visual device comprising: means for counting the number; means for holding the number of arbitrary objects; and means for geometrically analyzing the arbitrary object formation edge information image.
- the features of the present invention are as follows.
- First, a series of processes for generating the edge information of the arbitrary object from the moving image and normalizing the arbitrary object region separated from the background using the edge information are all realized by local processing. This allows the visual device to generate the arbitrary-object normalized image at high speed without depending on the shape and size of the arbitrary object, so that the application range of the visual device is wide. Therefore, the problems relating to the extraction and normalization of the arbitrary object are suitably solved.
- Second, a series of processes for generating the edge information of the arbitrary object from the moving image and normalizing the arbitrary object region separated from the background using the edge information are all realized by local processing. You.
- the visual device can quickly generate the arbitrary object normalized image without depending on the shape and size of the arbitrary object.
- the arbitrary object normalized image does not include the background, the arbitrary object
- the method of recognizing a normalized image is less affected by the position and the size of the background and the arbitrary object region than in the case of recognizing the arbitrary object region surrounded by the background.
- the range of applications for visual devices is wide. Therefore, the various problems relating to the recognition of the arbitrary object are suitably solved.
- environment data representing the type of the arbitrary object and the position in the environmental coordinates is created from the result of recognition of the arbitrary object normalized image and the orientation and the magnification of the moving camera.
- the visual device can create the environmental map, which is a collection of the environmental data.
- the environment map includes the environment data for a certain period, the distribution and the movement status of each type of the arbitrary object can be recorded, and the application range of the visual device is wide. Therefore, various problems concerning the recording of the arbitrary object can be suitably solved.
- the position of the arbitrary object that has not yet been recognized in the environment coordinates can be obtained.
- the application range of the visual device is wide. Therefore, the problems relating to the search for the arbitrary object are suitably solved.
- the application range of the visual device is wide. Therefore, the various problems relating to the counting of the arbitrary objects are suitably solved.
- the recognizing means can more quickly and accurately recognize the arbitrary object normalized image.
- the means for recognizing the arbitrary object normalized image generates the environment map before generating the recognition result. Therefore, the application range of the visual device is wide. Therefore, the problems relating to the recognition, search, and counting of the arbitrary object are suitably solved.
- a means for initializing the array operation unit and an input should be input.
- Visual device That is, this is an implementation form of an algorithm for realizing the digital image vibration function provided by the array operation unit by digital technology.
- the digital image is appropriately input in pixel units. Then, from the vibration of each band pixel value of the digital image to the output of each band pixel value of the vibration image is sequentially performed, and the process is repeated until the digital image is no longer input.
- a general-purpose processor can be used, and the parameter can be easily corrected.
- the array operation unit is initialized for each of the array operation units arranged in a lattice in a data processing device that implements a means for generating a coarse edge information image from a digital image.
- Means means for terminating the process if there is no digital image to be input, means for inputting each band pixel value of the digital image, and smoothing by smoothing each band pixel value of the digital image Means for generating each band pixel value of the image; means for taking the logarithm of each band pixel value of the smoothed image to generate each band pixel value of the logarithmically converted image; and each of the bands of the logarithmically converted image.
- this is an implementation form of an algorithm for realizing the generation function of the coarse edge information image provided by the array operation unit by digital technology.
- the digital image is appropriately input in pixel units.
- the smoothing of the digital image to the output of each band pixel value of the rough edge information image are sequentially performed, and the process is repeated until the digital image is not input.
- a general-purpose processor can be used, and the parameter can be easily corrected. It is not always necessary for the array operation unit to strictly wait for reception of neighboring pixels of the various images transmitted from the adjacent array operation unit.
- the array operation unit waiting for reception substitutes the band pixel value of its own. Because you can. At this time, there is a possibility that some noise may be added to the pixels of the various images generated by the array operation unit, but in the present invention, most of the noise is absorbed by each of the means.
- the edge processing and the timeout processing can be simultaneously and easily realized.
- the invention according to claim 8 is characterized in that, in the data processing device for realizing the means for generating the formed edge information image from the coarse edge information image, each of the array operation units arranged in a lattice pattern is arranged.
- Means for initializing the column operation unit, means for terminating the process if there is no digital image to be input or the coarse edge information 3 ⁇ 4 image, and each band pixel value of the digital image and the coarse edge information image Means for inputting band pixel values; means for separating the band pixel values of the digital image from the band pixel values of the coarse edge information image; and smoothing the band pixel values of the digital image.
- Means for generating a band pixel value of the maximum zero point image means for inverting the band pixel value of the maximum zero point image to generate a band pixel value of the basic edge information image; Means for shaping the band pixel value of the rough edge information image so as to approach the band pixel value of the above, and the band pixel value of the formed edge information image by interpolating the line width of the band pixel value of the rough edge information image. And a means for outputting the band pixel value of the formed edge information image. That is, this is an implementation form of an algorithm for realizing, by digital technology, the function of generating the formation edge information image provided by the array operation unit.
- the digital image and the rough edge information image are displayed.
- Input as appropriate in pixel units sequentially perform the steps from separation of the digital image and the coarse edge information image to output of each band pixel value of the formed edge information image, until the digital image and the coarse edge information image are no longer input. repeat.
- a general-purpose processor can be used, and the parameter can be easily corrected. It is not necessary for the array operation unit to strictly wait for reception of neighboring pixels of the various images transmitted from the array operation unit in the vicinity.
- the neighboring image of the various images is This is because, when the element cannot be received, the array operation unit waiting for reception can substitute its own band pixel value. At this time, there is a possibility that some noise may be added to the pixels of the various images generated by the array operation unit, but in the present invention, most of the noise is absorbed by each of the means.
- the edge processing and the time-out processing are simultaneously and simply realized by means of substituting the band pixel value of its own.
- the invention according to claim 9 is a means for initializing the array operation unit for each of the array operation units arranged in a lattice in the data processing device for realizing the means for detecting the position and size of the object area.
- Means for terminating the process if there is no rough edge information image to be input; means for inputting the band pixel value of the rough edge information image; and the band of the rough edge information image Means for converting to a pixel value, means for converting the amount of movement calculated from the overlapping information image into a band pixel value of the moving amount image, and means for moving the overlapping information image at a moving position indicated by the band pixel value of the moving amount image.
- the processes from the conversion to the overlapping information image to the output of each band pixel value of the overlapping information image are sequentially performed, and the process is repeated until the rough edge information image is no longer input.
- a general-purpose processor can be used, and the parameter can be easily corrected.
- the array operation unit does not necessarily have to strictly wait for reception of pixels in the vicinity of various images transmitted from the array operation unit in the vicinity. This is because when it is not possible to receive the neighboring pixels of various images from the neighboring array operation unit, This is because the array operation unit waiting for reception can substitute a pixel value corresponding to 0.
- most of the noise is absorbed by each of the means. is there.
- a formed edge information image can be input instead of the rough edge information image.
- the array operation units are arranged in a grid pattern, and the array operation unit is placed in the vicinity of the array operation unit.
- the formed edge information image is appropriately input in pixel units, and each band of the overlapping information image is converted from the conversion to the overlapping information image.
- the process up to the output of the pixel value is sequentially performed, and the process is repeated until the formation edge information image is no longer input.
- a general-purpose processor can be used, and the parameter can be easily corrected.
- the array operation unit It is not always necessary for the array operation unit to strictly wait for reception of neighboring pixels of the various images transmitted from the nearby array operation unit. This is because, when it is not possible to receive the neighboring pixels of the various images from the neighboring array operation unit, the array operation unit waiting to receive receives the band pixel value corresponding to 0. This is because it can be substituted. At this time, there is a possibility that some noise may be added to the pixels of the various images generated by the array operation unit, but in the present invention, most of the noise is absorbed by each of the means. . By means of substituting the band pixel value corresponding to 0, the edge processing and the time-out processing are simultaneously and easily realized.
- the object region image can be input instead of the rough edge information image.
- the array operation units are arranged in a grid pattern, and the array operation units are arranged close to each other.
- the object region image is appropriately input in pixel units, and the conversion from the overlapping information image to the The processing is sequentially performed until the output of each band pixel value of the overlapping information image, and is repeated until the object area image is no longer input.
- a general-purpose processor can be used, and the parameter can be easily corrected.
- the array operation unit It is not always necessary for the array operation unit to strictly wait for reception of neighboring pixels of the various images transmitted from the nearby array operation unit. This is because, when the neighboring pixels of various images cannot be received from the neighboring array computing unit, the array computing unit waiting for reception substitutes the band pixel value corresponding to 0. Because you can. At this time, although there is a possibility that some noise may be added to the pixels of the various images generated by the array operation unit, in the present invention, most of the noise is absorbed by each of the means. By means of substituting the band pixel value corresponding to 0, the edge processing and the timeout processing can be simultaneously and easily realized.
- a tenth aspect of the present invention is a data processing device for realizing a means for normalizing an object region, a means for initializing the array operation unit for each of the array operation units arranged in a grid, Means for ending the processing if there is no object area image or digital image to be processed; means for inputting the band pixel value of the object area image and each band pixel value of the digital image; and the band pixel value of the object area image. Means for separating the respective band pixel values of the digital image to generate band pixel values of the updated object region image and respective band pixel values of the updated image; and calculating a moving amount calculated from the updated object region image as a moving amount image.
- It is a visual device characterized by comprising: That is, this is an implementation form of the algorithm provided by the array operation unit for realizing the function of generating the normalized image by digital technology. After arranging the array operation unit in a lattice shape, coupling the array operation unit to each other near each other, and setting initial values of parameters of the array operation unit, the object region image and the digital image are converted into pixels.
- the input is appropriately performed in units, and the processes from separation of the object region image and the digital image to output of each band pixel value of the normalized image are sequentially performed, and the process is repeated until the object region image and the digital image are not input.
- a general-purpose processor can be used, and the parameter can be easily modified. It is not always necessary for the array operation unit to strictly wait for reception of neighboring pixels of various images transmitted from the adjacent array operation unit. This is because, when the neighboring pixels of the various images cannot be received from the neighboring array operation unit, the array operation unit waiting for reception substitutes a pixel value corresponding to 0. Because you can do it.
- the array operation units are arranged in a grid pattern for each of the array operation units arranged in a grid pattern in the data processing device that realizes pattern matching among the means for recognizing the normalized image.
- Terminating means means for inputting a band pixel value of the normalized image, means for calculating a matching result, means for updating a matching result image, means for outputting a band pixel value of the matching result image, It is a visual device characterized by having the following.
- this is an implementation form of the algorithm for realizing the pattern matching provided by the array operation unit in the image recognition means by digital technology.
- Arranging the array operation units in a grid pattern After setting the initial values of the parameters of the array operation unit, the template image and the normalized image are appropriately input in pixel units, and the matching result is calculated based on the calculation of the matching result. The output of each band pixel value is sequentially performed, and the process is repeated until the normalized image is no longer input.
- a general-purpose processor can be used, and the above parameters can be easily corrected. It is not always necessary for the array operation unit to strictly wait for reception of neighboring pixels of the various images transmitted from the adjacent array operation unit.
- the array computing unit waiting for reception substitutes its own band pixel value. Because you can. At this time, there is a possibility that some noise may be added to the pixels of the various images generated by the array operation unit, but in the present invention, most of the noise is absorbed by each of the means. By means of substituting the band pixel value of its own, the edge processing and the timeout processing can be realized simultaneously and simply.
- the invention according to claim 12 is characterized in that, for each of the array operation units arranged in a lattice in the data processing device, which implements means for separating an object region using a formed edge information image, the array operation unit is arranged in a lattice. Means for connecting the non-linear oscillator in the array operation unit, the non-linear oscillator in the vicinity of the non-linear oscillator described above with a coupling value, and means for initializing the array operation unit.
- the invention of claim 13 is means for inputting data, means for sequentially storing the data, means for transferring the data between the array operation units, and means for calculating using the data, Means for outputting the data; means for arranging the array operation units in a grid pattern with respect to the array operation unit having; and means for interconnecting neighbors from each other based on the positional relationship between the array operation units.
- a visual communication device comprising: means for communicating the data between adjacent array operation units; and means for operating each of the array operation units independently.
- each of the array operation units has the same operation step regardless of where it is arranged in a grid.
- the same circuits that implement the array operation unit can be regularly arranged on a plane, and only those adjacent circuits need to be connected. Therefore, the wiring amount is small, and the number of circuits can be increased or decreased according to the size of the image to be handled. And each circuit can be operated in parallel.
- the array operation unit is implemented by software, a visual device in which the array operation units are arranged in a lattice can be executed by a program with high parallelism.
- the array operation unit comprises: a processor having means for processing the input data; a memory for storing a program for processing the data and a variable; A controller for communicating with an array operation unit, wherein the controller stores the input data in the memory, and transmits the variable in the memory to an adjacent array operation unit.
- the array operation unit can use a general-purpose processor for processing the input data and a general-purpose memory for storing a program and variables for processing the data. When the array operation unit is interconnected with at most four neighbors, the controller only needs to transmit a variable to the adjacent array operation unit only.
- the variables of the array operation unit that are not included in the neighborhood 4 are once transmitted to the adjacent array operation unit, so that the variables are received by having the variable transmitted to itself again. be able to.
- the means for transmitting each of the variables in the memory to the adjacent array operation unit may transmit the own variable to the array operation unit which is not included in the vicinity of four. This allows the array operation unit to communicate appropriate data with the array operation unit having eight or more neighborhoods, even though the array operation unit is connected to only the adjacent array operation unit as hardware. Further, of the above problems, various problems relating to hardware implementation and real-time processing are suitably solved. BRIEF DESCRIPTION OF THE FIGURES
- FIG. 1 is a block diagram of a visual device for counting the number of moving objects from a frame image.
- FIG. 2 is a block diagram of a visual device that counts the number of moving objects from a frame image using an object Z background separating means.
- FIG. 3 is a block diagram of a visual device for counting the total number of objects from a frame image.
- FIG. 4 is a block diagram of a visual device that counts the total number of objects from a frame image using an object background separation unit.
- FIG. 5 is a block diagram of a visual device for calculating the ratio between a moving object and a stationary object.
- FIG. 6 is a block diagram of a visual device for controlling a moving camera.
- FIG. 7 is a block diagram of a visual device that generates a normalized image of an object.
- FIG. 8 is a block diagram of a visual device that outputs an object recognition result.
- FIG. 9 is a block diagram of a visual device for generating an environmental map.
- FIG. 10 is a block diagram of a visual device that controls a mobile camera using an environment map.
- FIG. 11 is a block diagram of a visual device for counting the number of arbitrary objects.
- FIG. 12 is a block diagram of a visual device speeded up by geometric analysis.
- FIG. 13 is a block diagram in which array operation units are arranged in a grid.
- FIG. 14 is a flowchart showing the algorithm of the image storage means of the present embodiment.
- FIG. 15 is a flowchart showing an algorithm of the image vibration means of the present embodiment.
- FIG. 16 is a flowchart showing an algorithm of the edge information generating means of the present embodiment.
- FIG. 17 is an explanatory diagram of a case where coarse edge information is formed into formed edge information using a digital image.
- FIG. 18 is a flowchart showing an algorithm of the edge information forming means of the present embodiment.
- FIG. 19 is an explanatory diagram of a case where low-resolution coarse edge information generated from a low-resolution digital image is formed as formed edge information.
- FIG. 20 is an explanatory diagram of a case where an area of low-resolution coarse edge information generated from a low-resolution digital image is cut out and then formed as formed edge information.
- FIG. 21 is an explanatory diagram for detecting the position and size of an object in an edge information image.
- FIG. 21 is an explanatory diagram for detecting the position and size of an object in an edge information image.
- FIG. 22 is a flowchart showing an algorithm of the position / size detecting means of the present embodiment.
- FIG. 23 is an explanatory diagram for detecting the position and size of an object in an object area image.
- FIG. 24 is an explanatory diagram in the case of normalizing the cut-out area of the digital image.
- FIG. 25 is a flowchart showing an algorithm of the area normalizing means of the present embodiment.
- Fig. 26 shows the normalized image of this embodiment. It is a flowchart which shows the algorithm of a holding means.
- FIG. 27 is an explanatory diagram of a case where pattern matching is performed on a normalized image from among template images.
- FIG. 28 is a flowchart showing an algorithm of pattern matching in the image recognition means of the present embodiment.
- FIG. 29 is an explanatory diagram showing a state where edge information of a triangle is separated into an inner area and an outer area of the triangle.
- FIG. 30 is a flowchart showing an algorithm of the object / background separation means of the present embodiment.
- FIG. 31 is an explanatory diagram showing a state where edge information of a triangle in a broken line state is separated into an inner region and an outer region of the broken line triangle.
- FIG. 32 is an explanatory diagram showing a state where edge information obtained by overlapping two triangles is separated into two triangle regions and a background region.
- FIG. 33 is an explanatory diagram showing a state where edge information in a dashed state when two circular object regions are overlapped is separated into two circular regions and a background region.
- FIG. 34 is a block diagram of the internal structure of the array operation unit.
- FIG. 35 is a block diagram of the controller.
- FIG. 36 is an explanatory diagram showing input / output signals of the flag decoder.
- FIG. 37 is an explanatory diagram showing input / output signals of the flag encoder.
- FIG. 38 is a flowchart showing an algorithm in which the processor transmits data to the adjacent array operation unit via the controller.
- FIG. 39 is a flow chart showing an algorithm in which a controller receives data from an adjacent array operation unit.
- FIG. 40 is a flowchart showing an algorithm in which the processor receives data from the upper input register.
- FIGS. 1 to 12 include an image acquisition unit 11 (see FIG. 1) that receives an image signal (frame image 1) of a video camera and converts it into a digital image 11 1 having an appropriate format and size; Image storage means 1 2 (see FIG. 1) for storing 11 for a certain period of time; image vibration means 13 (see FIG. 3) for vibrating the digital image 11 1 using a digital circuit; Edge information generating means 14 (see FIGS. 1 and 3) for generating coarse edge information 1 12 of moving object 2 or stationary object 3 from image 11 1 1 and coarse edge information 1 12 more accurately. And clear formation Edge information forming means 15 (see FIG. 1) (see FIG. 1) that receives an image signal (frame image 1) of a video camera and converts it into a digital image 11 1 having an appropriate format and size; Image storage means 1 2 (see FIG. 1) for storing 11 for a certain period of time; image vibration means 13 (see FIG. 3) for vibrating the digital image 11 1 using a digital circuit; Edge information generating means 14 (see FIGS. 1 and 3)
- edge information 1 14 1) formed on edge information 1 14; object Z background separating means 16 (see FIG. 2) separating an area divided by formed edge information 1 14; A position size detecting means 17 (see FIGS. 1 and 2) for detecting the position and size of each area divided or separated by the formed edge information 114, and an appropriate size.
- Pixel counting means 18 for counting the number of areas from the position of a certain area (see Fig. 1) and pixel number holding means 19 for outputting the number of areas or a percentage of the number of areas (see Figs. 1 and 5) ) Etc., and will be described with reference to the drawings.
- a moving object counting unit 101 inputs a frame image 1 of a moving image captured by a video camera to an image obtaining unit 11, and stores an image storage unit 12 and edge information generation.
- the frame image 1 may be a still image continuously shot by a digital camera.
- the image acquisition means 11 inputs a frame image 1 of a moving image from a video camera
- the frame image 1 is converted into a digital signal by AZD conversion using a general capture board. Convert to digital image 1 1 1
- a voltage of a CCD image sensor or the like can be directly input, it is only necessary to convert the digital signal into an appropriate number of bits by AZD conversion.
- the moving image is a digital signal, decompress it if it is compressed, or input it as is if it is not.
- an arbitrary frame image 1 can be cut out of the moving image, so that the frame image 1 is cut out to be a digital image 111.
- the image acquisition means 11 converts the image data into a format in which the image data can be referenced in pixel units, and a moving object counting unit.
- the required image size is cut out in 101 and output as digital image 111. If the image acquisition unit 1.1 can output all the pixels of the digital image 11 1 in parallel, the communication from the image acquisition unit 11 to the image storage unit 12 can be performed in parallel for each pixel. it can.
- the moving object The digital image 111 is stored for a certain period of time in accordance with the time resolution of the counting unit 101 or the calculation capability of each means. In other words, even if the digital image 11 1 is input during this fixed time, the image storage means 12 does not change the stored image, so that each subsequent means inputs the same digital image 11 1 at a different timing. be able to. Moreover, since the image storage means 12 does not perform image processing on the digital image 111, all the pixels of the digital image 111 are stored while maintaining the two-dimensional phase relationship. If the image storage means 12 can output all the pixels of the digital image 11 in parallel, the communication from the image storage means 12 to the edge information generation means 14 can be performed in parallel for each pixel. it can.
- the edge information generating means 14 When the edge information generating means 14 receives the digital image 1 11 from the image storing means 12, the edge information generating means 14 compares the digital image 1 1 1 with the previously input digital image 1 1 1 to obtain the coarse edge information image 1 1 3 Generate. Since the edge information generating means 14 can generate the coarse edge information image 113 only by the neighborhood processing for each pixel, it is suitable for parallelization. If the edge information generating means 14 can output all the pixels of the coarse edge information image 113 in parallel, the communication from the edge information generating means 14 to the edge information forming means 15 is performed for each pixel. Can be done in parallel.
- the edge information forming means 15 receives the coarse edge information image 113 from the edge information generating means 14, the coarse edge information image is referred to by referring to the digital image 111 stored in the image storage means 122. 1 1 3 A more accurate and clear formed edge information image 1 15 of the moving object 2 is generated. Since the edge information forming means 15 can generate the formed edge information image 115 only by the neighborhood processing for each pixel, it is suitable for parallelization. If the edge information forming means 15 can output all the pixels of the formed edge information image 115 in parallel, the communication from the edge information forming means 15 to the position / size detecting means 17 is performed for each pixel. Can be done in parallel.
- the position and size detecting means 17 When the position and size detecting means 17 inputs the formed edge information image 1 15 from the edge information forming means 15, the position and size of the area of the moving object 2 indicated by the formed edge information 114 are detected. I do. Since the position / size detection means 17 can generate a duplicate information image 13 2 representing the detection result of the position and size of the area of the moving object 2 by only neighborhood processing for each pixel, it is suitable for parallelization. ing. If the position / size detection means 17 is the same for all pixels of the Can be output in parallel, the communication from the position / size detecting means 17 to the pixel counting means 18 can be performed in parallel for each pixel.
- the pixel counting unit 18 When the pixel counting unit 18 receives the overlapping information image 13 2 from the position Z size detecting unit 17, the pixel counting unit 18 counts the number of pixels representing the position of the area of the moving object 2 having an appropriate size. It is more convenient to perform this process sequentially instead of in parallel. Therefore, each pixel of the overlapping information image 132 is stored in the linear memory. The number of pixels representing the position of the area of the moving object 2 is output from the pixel counting means 18 to the pixel number holding means 19.
- the moving object counting unit 101 can output the number of moving objects.
- the moving object counting unit 101 can have real-time properties. Therefore, it is suitable for counting the number of fast moving objects or for processing a large amount of moving images in a short time. Also, if some means are implemented by software on a single processor, the computational speed can be reduced and the cost can be reduced. Therefore, it is suitable for applications where it may take more than a few minutes for the counting result to be obtained.
- the position Z size detecting means 17 detects the position and size of the area of the moving object 2 using the formed edge information image 1 15 generated by the edge information forming means 15, Depending on the density, the number of pixels representing the position of the area of the moving object 2 may be different from the number of the moving objects 2.
- the main causes are that the formed edge information 114 does not always accurately extract the edge of the moving object 2 and that the position / size detection means 17 uses the formed edge information 111 to detect the moving object 2 It is not to determine the shape. Therefore, when the density of the moving object 2 increases, formed edge information 114 generated from different moving objects 2 may be combined and confused with edge information of a non-existent object.
- the shape of the moving object 2 should be determined from the formed edge information 1 1 4 .
- the conventional geometric analysis method requires a large amount of calculation due to global processing, and the accuracy of the determination result is high. Increases the computation time exponentially. Therefore, as a means for solving this problem by neighborhood processing, the object Z background separation means 16 (see FIG. 2) can be used.
- the object / background separating means 16 separates the pixels included in the object area 141 and the pixels included in the background into different groups. The separation results are output sequentially in groups. If the object regions 14 1 are clearly distinguished by the formed edge information 114 even though they are adjacent to each other, the object Z background separation means 16 can separate these object regions into different groups. . Therefore, the number of groups can be more than two.
- the object / background separation means 16 is suitable for parallelization since the object area 141 and the background can be separated only by neighborhood processing for each pixel. If the object Z background separating means 16 can output all the pixels of the object area image 14 2 in parallel, the communication from the object background separating means 16 to the position / size detecting means 17 is performed for each pixel. Can be performed in parallel.
- the object background separation means 16 by using the object background separation means 16, it is possible to separate the area of the moving object 2 from other background areas while supplementing the formed edge information 114 only by neighborhood processing. it can.
- the position / size detecting means 17 extracts only the area of a certain size and extracts the position. Can be specified. Therefore, if it is known in advance that the moving object 2 resembles a circle, the pixel counting means 18 can use the object / background separation means 16 to determine the area of the moving object 2 having a specific size. The number of pixels representing the position can be counted. That is, the moving object counting unit 101 can output the moving object number with higher accuracy.
- moving object counting section 101 operates when counting the number of moving objects 2 in two or more frame images 1 of a moving image.
- the total number of moving objects 2 and still objects 3 in one frame image 1 of a moving image that is, the total number of objects is counted.
- the basic operation is the same as the case of counting the number of moving objects 2 using the frame image 1 of the moving image described above. That is, the stationary object 3 in the frame image 1 may be apparently replaced with the moving object 2 in the frame image 1 by an appropriate method. Then the rough object 3 of the stationary object 3 Since the object information 112 is generated, the moving object counting unit 101 can also count the number of the stationary objects 3.
- the simplest way to imitate the stationary object 3 as the moving object 2 is to finely vibrate the video camera (digital camera) or the stationary object 3 itself using a shaking table or the like.
- video camera digital camera
- stationary object 3 itself using a shaking table or the like.
- image vibrating means 13 is a means that does not use a physical mechanism.
- the image vibration unit 13 When the digital image 1 11 is input from the image acquisition unit 11, the image vibration unit 13 simultaneously transmits the digital image 1 1 1 in units of an image so that the stationary object 3 vibrates up and down and left and right within a range of about 3 pixels. , Or move individually in pixel units. If the image oscillating means 13 can output all the pixels of the digital image 11 1 in parallel, the communication from the image oscillating means 13 to the edge information generating means 14 can be performed in parallel for each pixel. it can.
- the image vibrating means 13 vibrates the stationary object 3 in the digital image 11 generated by the image acquiring means 11 up, down, left and right within a range of about 3 pixels.
- the edge information generating means 14 can regard the stationary object 3 as the moving object 2 and generate the coarse edge information 1 1 2 of the stationary object 3.
- the object Z background separating means 16 separates the object area 14 1 and the background area only by the forming edge information 114. Therefore, if it is known in advance that the moving object 2 and the stationary object 3 are similar to a circle, as shown in FIG. 4, the formed edge information image 115 generated by the edge information forming means 15 is obtained.
- the position / size detection means 17 can count the number of pixels representing the position of the area of the moving object 2 and the stationary object 3 of a specific size. That is, the total object counting unit 102 can output a more accurate total object number.
- the visual device uses the whole object counting unit 102 provided with almost the same means as the moving object counting unit 101 that counts the number of moving objects 2 in a moving image, Moving image files The total number of moving objects 2 and stationary objects 3 in the frame image 1 can be counted. Therefore, consider a method of counting the number of moving objects 2 and the number of stationary objects 3 in parallel using one visual device. At this time, if the number of moving objects 2 and stationary objects 3 can be determined, it is very easy to calculate the ratio of moving objects 2 or the ratio of stationary objects 3 to all objects. Therefore, a visual device for calculating the ratio between the moving object 2 and the stationary object 3 will also be described.
- the overall configuration of the visual device that calculates the number of moving objects 2 and all objects and calculates the ratio of moving objects 2 and stationary objects 3 is roughly divided into a moving object counting unit 101 and It consists of a total object counting unit 102.
- the moving object counting unit 101 and the total object counting unit 102 each include the above-described means. Note that since the image acquisition means 11 and the image storage means 12 of the moving object counting section 101 and the total object counting section 102 operate exactly the same, FIG. 5 shows the image acquisition means 11 and the image storage means.
- the number-of-pixels holding unit 19 of the moving object counting unit 101 inputs the total number of objects from the pixel counting unit 18 of the total object counting unit 102, and
- the number-of-pixels holding means 19 inputs the number of moving objects from the pixel counting means 18 of the moving-object counting unit 101 to determine the ratio between the moving object 2 and the stationary object 3 by an external signal such as a ratio switching signal. That is, it can be switched to calculate and convert to floating point notation.
- a ratio switching signal such as a ratio switching signal
- the corresponding functions may be added to the pixel number holding means 19.
- the visual device can freely make such a change depending on the application.
- the visual device described in claims 1 to 3 is configured by the moving object counting unit 101 and the total object counting unit 102.
- This is a device for counting the number of moving objects 2 and stationary objects 3 in a captured frame image 1. Therefore, if the moving object 2 and the stationary object 3 are included in the frame image 1, these visual devices can count the number of the moving object 2 and the stationary object 3. However, once the moving object 2 and the stationary object 3 deviate from the frame image 1, these visual devices And the number of stationary objects 3 cannot be counted. Therefore, in the following, a visual device according to claims 4 and 5, in which the number is counted while constantly searching for the moving object 2 and the stationary object 3 using the moving camera 10 (see FIG. 6), will be described.
- the mobile camera 10 has a mechanism for moving in the horizontal direction and the vertical direction, and can perform pan and tilt, respectively, by inputting a command for controlling the movement angle from the outside. It is also assumed that the mobile camera 10 has a mechanism for changing the magnification of a captured image, and can perform zooming by inputting a command for controlling the magnification from outside. Therefore, the moving force camera 10 can finely vibrate the camera itself by an external command. As a result, the frame image 1 of the moving image captured by the moving camera 10 is blurred, and the object in the frame image 1 is captured as if it were vibrating.
- the moving camera 10 when the moving camera 10 is used, the area of the whole object can be counted only by the moving object counting unit 101 without using the image vibration means 13 of the whole object counting unit 102.
- This method has problems in processing speed and counting accuracy because it uses the physical mechanism of the moving camera 10.However, only the moving object counting unit 101 can cover both areas of the moving object 2 and all objects. Since it can be counted, it is suitable for use in counting moving objects 2 and stationary objects 3 in a wide place that cannot be accommodated by the frame image 1.
- the mobile camera 10 can output the current position moved by a movement command such as pan, tilt, and zoom as necessary, and further, whether the mobile camera 10 is currently moving or stopped. It is assumed that the status can be output as needed.
- the visual device can control the pan, tilt, and zoom of the moving camera 10
- the moving device 10 always moves the moving object 2 and the stationary object 3 to an appropriate size. The direction and the magnification of the moving camera 10 can be changed so that a photograph can be taken.
- FIG. 6 shows a visual device according to claim 4 in which the moving camera 10 is provided with basic means for constantly photographing the moving object 2 and the stationary object 3 with an appropriate size.
- the visual device uses three coordinate systems, that is, a camera coordinate system, an image coordinate system, and an environment coordinate system, depending on the function.
- the camera coordinate system is a three-dimensional spherical coordinate system inside the camera that the moving camera literally uses to control pan, tilt and zoom in each of the minimum control units.
- the origin of the camera coordinate system is a position unique to the moving camera called a home position.
- the camera coordinate system is the only coordinate system that can represent the physical positions of the moving object 2 and the stationary object 3.
- the image coordinate system is a two-dimensional coordinate system in which the center is the center of the frame image 1 captured by the moving camera 10 and has a pixel unit. This is used to indicate at which pixel in the frame image 1 the moving object 2 and the stationary object 3 are located. Therefore, the image coordinate system is suitable for distinguishing the fine positions of multiple objects in the frame image 1, but the image coordinate system alone can represent the physical positions of the moving object 2 and the stationary object 3. Can not.
- the environment coordinate system is a three-dimensional spherical coordinate system used by the visual device to logically unify the positions of the moving object 2 and the stationary object 3 inside.
- the environment coordinate system uses angles in radians in the horizontal and vertical directions, and uses real numbers in units of 1.0 in the distance direction to represent the product of the size of the object and the distance to the object. In general, since the size of the object does not change extremely, the distance to the object and the magnification of the moving lens 10 may be regarded as proportional.
- the origin of the environment coordinate system is arbitrary. In other words, the environment coordinate system is used to represent the relative coordinates of any two points on the environment coordinate system in principle.
- the visual device can distinguish a plurality of objects by projecting objects in the environment that can be photographed by the moving camera 10 onto the environment coordinate system.
- the camera coordinate system and the image coordinate system need to be mutually coordinate-transformed with the environment coordinate system.
- Means that fulfill that role are camera / environmental coordinate conversion means 20, image environment coordinate conversion means 21, and motion control means 23. These means obtain the respective units of the camera coordinate system and the image coordinate system from the specifications of the moving force camera 10 and the image acquisition means 11 and calculate a matrix for conversion into the environment coordinate system. By calculating the inverse of the transformation matrix from the camera coordinate system to the environment coordinate system, the transformation matrix from the environment coordinate system to the camera coordinate system can also be obtained. However, since the origin of the camera coordinate system is the home position of the moving camera 10, the position of the environment coordinate system converted from the camera coordinate system is relative to the moving camera 10's home position on the environment coordinate system.
- the position of the camera coordinate system converted from the environment coordinate system is It is a relative position from the current position of the moving camera 10 on the system.
- the image coordinate system is a two-dimensional coordinate system, it cannot be converted to the environment coordinate system using only the image coordinate system. Therefore, the image environment coordinate conversion means 21 uses the orientation and magnification of the moving camera 10 expressed in the environment coordinate system and the size of the area of the moving object 2 and the stationary object 3 in the frame image 1 as needed to perform the conversion. By calculating the matrix, it is possible to convert from the image coordinate system to the environment coordinate system.
- the position of the environment coordinate system converted from the image coordinate system is a relative position from the center of the image.
- the remaining means of the visual device according to claim 4 can be used to move the moving object 2 and the stationary object 3 represented by these three coordinate systems in order to always photograph the moving object 2 and the stationary object 3 at an appropriate size. It can be considered as a means for generating and transforming the position of the object 3. Therefore, Fig. 6 is explained based on the camera coordinate system, image coordinate system and environment coordinate system.
- the image acquisition unit 11, the edge information generation unit 14 and the position / size detection unit 17 are composed of a moving object counting unit 101 and a total object counting unit 102. This is the same as that described for the visual device described. However, in the visual device described in claims 1 to 3, the formed edge information image 1 15 generated by the edge information forming means 15 is input to the position size detecting means 17. In the visual device according to the fourth aspect, the coarse edge information image 113 generated by the edge information generating means 14 is input to the position / size detecting means 17. Of course, in this visual device, it is also possible to input the formed edge information image 115 to the position / size detecting means 17 by using the edge information forming means 15, but here, for the following reasons.
- the performance of this visual device is sufficient without using the edge information forming means 15.
- the visual device does not need to count the number of areas of the moving object 2 and the stationary object 3 as required by the visual device described in claims 1 to 3. Rather, it is important that this visual device searches for the moving object 2 and the stationary object 3 and matches the direction and the magnification of the moving camera 10 to the direction.
- the position Z size detecting means 17 can obtain the approximate size of the area of the moving object 2 and the stationary object 3 even from the rough edge information 112.
- the edge information forming unit 15 generates the formed edge information image 115 using the rough edge information image 113 generated by the edge information generating unit 14. That is, the edge information forming means 1 The moving object 2 moves while 5 generates the formed edge information image 1 15.
- the moving speed of the moving camera 10 is not so fast because the moving camera 10 has a physical mechanism. Therefore, if the visual device moves the moving camera 10, the control of the moving camera 10 may not be able to keep up depending on the moving speed of the moving object 2. Based on these facts, the visual device according to claim 4 does not necessarily need accurate and clear formation edge information 114 of the moving object 2 and the stationary object 3.
- the force-melano environment coordinate conversion means 20 and the image Z environment coordinate conversion means 21 include a moving object 2 and a stationary object 3 represented by the overlap information image 13 2 generated by the position and size detection means 17. Converts the position of the area to the position in the environment coordinate system. At this time, if the total number of positions of the areas of the moving object 2 and the stationary object 3 in the frame image 1 is 2 or more, two or more positions exist on the environment coordinate system. Therefore, in order to control the pan, tilt, and zoom of the moving camera 10 and adjust the direction and magnification of the moving camera 10 to one of the objects, it is necessary to select one of the positions on the environment coordinate system. is there.
- the position selecting means 22 selects one position on the environment coordinate system according to a predetermined criterion.
- the criteria used here are mainly as follows. First, select the closest (or largest) object in the environment coordinate system. This means that if the object is far (or small), the edge information generating means 14 may have generated noise, so that the object with a high probability of being at a position on the environmental coordinate system even if it is a little. select. Second, if multiple locations are concentrated within a certain range on the environment coordinate system, select one of them. This has two possibilities. One is the possibility that the edge information generating means 14 has generated coarse edge information 1 1 2 dispersed for one object, and the other is the possibility that a plurality of objects actually exist.
- the position selecting means 22 can select one position on the environment coordinate system.
- the visual device according to claim 4 is required to pan, tilt, and zoom the mobile camera 10 to the following position in addition to the position selected by the position selecting means 22.
- this vision The device must vibrate the moving camera 10. Therefore, in order to vibrate the mobile camera 10, the vibration command generation means 25 specifies a position where the mobile camera 10 moves as a position on the environment coordinate system.
- the position specified by the vibration command generation means 25 is determined by a pseudo random number or the like within a range where the moving camera 10 does not vibrate extremely.
- the visual device according to claim 4 is required to pan, tilt, and zoom the moving camera 10 by inputting a control command from outside. In a general use, panning, tilting and zooming are performed in the direction and magnification at which the moving camera 10 is currently facing. Therefore, the control command input means 24 is used to temporarily store the control command, and then store the control command. Calculate the position on the environment coordinate system with the current position of 0 as the origin. By improving the control command input means 24, it is possible to easily move the mobile camera 10 to a specific position.
- the motion control means 23 selects one from the three positions on the environment coordinate system described above.
- the control command input means 24, the position selection means 22 and the vibration command generation means 25 are selected in the order of the input positions. .
- the selected position is converted from the environment coordinate system to the force coordinate system.
- the camera command generation means 26 replaces the command with a command recognizable by the mobile camera 10 and transmits the command to the mobile camera 10.
- the visual device according to claim 4 can control pan, tilt, and zoom of the moving camera 10.
- the state of the mobile camera 10 is determined.
- the motion control means 23 selects one of the three positions and outputs it to the camera command generation means 26, and then issues a command to the mobile camera 10 to inquire whether the mobile camera 10 is moving or not. It instructs the camera command generation means 26 to transmit, and waits until this information is received from the moving camera 10 via the camera Z environment coordinate conversion means 20. If the received information is moving, it instructs the camera command generation means 26 again to transmit a command to the mobile camera 10 for inquiring whether the mobile camera 10 is moving. If the information received is stopped, mobile camera 10 It instructs the camera command generation means 26 to transmit a command for inquiring the current direction and magnification to the mobile camera 10. During this period, the motion control means 23 does not select three positions.
- the camera command generation means 26 transmits a corresponding command to the mobile camera 10 according to the instruction from the motion control means 23.
- the camera / environment coordinate conversion means 20 sends the information as to whether or not the moving camera 10 is moving and the motion information to the movement control means 23 as it is. Convert to a coordinate system position.
- the visual device according to claim 4 can control pan, tilt, and zoom of the mobile camera 10 while sequentially examining the state of the mobile camera 10.
- the visual device according to claim 4 has been described in which the moving camera 10 is provided with basic means for constantly photographing the moving object 2 and the stationary object 3 in the frame image 1 at an appropriate size.
- the moving object 2 and the stationary object 3 are not always present in the range where the moving camera 10 is shooting, and above all, the moving object 2 moves to another position from the range where the moving camera 10 is shooting. It is natural to think that it will.
- the visual device described in claim 5 described below is different from the visual device described in claim 4 in that the shape and color of the moving object 2 and the stationary object 3 in the frame image 1 are more accurately recognized.
- an edge information forming unit 15, an object Z background separating unit 16, a region normalizing unit 27, and a normalized image holding unit 28 are added to the visual device described in claim 4.
- a normalized image 144 is generated.
- the edge information forming means 15 and the object Z background separating means 16 comprise a moving object counting section 101 and a total object counting section 102, and the visual information described in any one of claims 1 to 3. This is the same as the edge information forming means 15 and the object Z background separating means 16 of the apparatus.
- the area normalizing means 27 receives the object area image 144 and the digital image 111 from the object background separating means 16 and the image acquiring means 111, respectively.
- a normalized image 1 45 is generated by cutting out from 11 and complementing and enlarging as much as possible according to the image size of the digital image 111 while deforming the separated object region 144.
- the region normalizing means 27 can normalize the separated object region 144 by only neighborhood processing for each pixel, and is suitable for parallelization. If the area normalizing means 27 can output all the pixels of the normalized image 144 in parallel, the communication from the area normalizing means 27 to the normalized image holding means 28 is performed in parallel for each pixel. Can be done.
- the output destination of the normalized image 144 requests the normalized image 144 of an appropriate format. If this is the case, convert the normalized image 145 to the format required by the output destination of the normalized image 145. Thereafter, the normalized image holding means 28 stores the normalized image 145 for a certain period of time until the normalized image 145 is reliably transmitted to the output destination of the normalized image 145. If the format to be converted is limited, the normalized image holding means 28 can convert the normalized image 145 only by neighborhood processing for each pixel, and is suitable for parallelization. If the normalized image holding means 28 can output all the pixels of the normalized image 144 in parallel, the communication from the normalized image holding means 28 to the output destination of the normalized image 144 is , Can be performed in parallel for each pixel.
- the visual device shown in FIG. 7 can provide a normalized image of the moving object 2 and the stationary object 3 as similar as possible. Can be generated.
- the output destination of the normalized image 1 45 needs to recognize the moving object 2 and the stationary object 3
- the output destination of the normalized image 1 45 is the moving object 2 and the stationary object 3 in the frame image 1.
- the edge information generating means 14, the edge information forming means 15, and the object It is not always necessary to make the resolution or image size of the image input by each means between the body / background separation means 16, the area normalization means 27 and the normalized image holding means 28.
- a low-resolution digital image 116 with a reduced resolution of the digital image 111 is input to the edge information generating means 14 while the edge information forming means 15 Input the coarse edge information image 1 13 obtained by enlarging the image size of the low resolution coarse edge information image 1 17 generated by the edge information generating means 14 to the digital image 11 1 size by an appropriate method.
- the load on the edge information generating means 14 can be reduced by inputting the digital image 111 to the object Z background separating means 16 and the area normalizing means 27.
- the pan / tilt / zoom control of the moving camera 10 after the position / size detection means 17 can be performed with little change in the quality of the normalized image 144 generated after the edge information forming means 15. Can be faster. Therefore, when this method is further advanced, the edge information forming unit 15 cuts out the area where the coarse edge information 1 12 exists from the low-resolution coarse edge information image 1 17 generated by the edge information generating unit 14.
- the extracted coarse edge information image 1 19 is input, and the object / background separating means 16 and the area normalizing means 27 are provided with an area at the same position as the cut coarse edge information image 1 19 from the digital image 111.
- the cutout digital image 120 obtained by cutting out the image, it is possible to reduce the load of generating the normalized image 144 after the edge information forming means 15.
- the moving camera 10 can photograph the moving object 2 and the stationary object 3 at an appropriate size in the center of the frame image 1 by the visual device according to claim 4, a rough cutout of the digital image 111
- the cutout areas of the edge information image 119 and the cutout digital image 120 can be determined in advance. According to this method, the visual device shown in FIG. 7 can achieve the same performance as the object search device using a wide-angle camera and a high-resolution camera by using one moving camera 10.
- the visual device shown in FIG. 8 generates a recognition result by adding an image recognition means 29 and a recognition result holding means 30 to the visual device shown in FIG.
- the image recognizing means 29 receives the normalized image 1 45 from the area normalizing means 27, and performs appropriate pattern recognition on the normalized areas 1 4 4 of the moving object 2 and the stationary object 3 in the normalized image 1 45. One Recognition using the method, and outputs the recognition result. Since the normalized image 1.45 input to the image recognition means 29 has the shape of the moving object 2 and the stationary object 3 deformed by the area normalization means 27, the image recognition means 29 has a stroke extraction method. Rather than performing geometric analysis using a method that is resistant to misalignment such as Fourier transform and Hough transform, it is more appropriate to perform pattern matching that compares the input image with the template image.
- a neural network such as a perceptron that can learn a template image by an error backpropagation method (back propagation) can be used.
- backpropagation back propagation
- parallelization and speeding up are possible by using the neural network exclusive filer.
- the recognition result of the normalized image 144 is output from the image recognition means 29 to the recognition result holding means 30.
- the recognition result holding means 30 When the recognition result holding means 30 inputs the recognition result of the normalized image 144 from the image recognition means 29, if the output destination of the recognition result requests a signal in an appropriate format, the recognition result is output. Convert the recognition result to the format required by the destination. Thereafter, the recognition result holding means 30 stores the recognition result for a certain period of time until the recognition result is reliably transmitted to the output destination of the recognition result.
- the visual device shown in FIG. 8 can recognize the recognition results of the moving object 2 and the stationary object 3 captured by the moving force camera 10 with an appropriate size. Can be generated.
- the output destination of the recognition result can use the visual device shown in FIG. 8 as a device for recognizing the moving object 2 and the stationary object 3 photographed by the moving camera 10.
- the visual device shown in FIG. 9 generates an environmental map by adding environmental understanding means 31, clocking means 32 and environmental map holding means 33 to the visual device shown in FIG.
- the timer means 32 outputs the current time in 1 millisecond units by a timer circuit.
- the current time is constantly output from the timing means 32 to the environmental understanding means 31.
- the environment understanding means 3 1 receives the recognition results of the moving object 2 and the stationary object 3 from the image recognition means 29, and outputs the recognition result, environment data including the position of the moving camera 10 on the environment coordinate system and the current time. create.
- the environment understanding means 3 1 inputs the positions of the areas of all moving objects 2 and stationary objects 3 in the frame image 1 on the environment coordinate system from the image Z environment coordinate conversion means 21. Then, the recognition result consisting of Null and the night, the position of the moving camera 10 on the environment coordinate system and the position of one moving object 2 and the stationary object 3 in the frame image 1 on the environment coordinate system are added.
- the environment data consisting of the position and the current time is created by the number of areas of the moving object 2 and the stationary object 3.
- the environment map is a set of environment data created a predetermined time before the current time, and the position in the environment data is represented by an environment coordinate system whose origin is the home position of the mobile camera 10.
- the environmental understanding means 31 adds and deletes environmental data as time elapses from the environmental map.
- the environmental understanding means 31 deletes the duplicated environmental data, and the position in the environment data where the recognition result is null data is near the position in other environmental data in which the recognition result is not null data. , Delete environmental data for which the recognition result is nulle overnight.
- the recognition result in these environment data is If matches, delete the former environmental data.
- the accuracy of the environmental map is determined by the recording time of the environmental data and the range near the position of the environmental data.
- the environmental map is output from the environmental understanding means 31 to the environmental map holding means 33.
- the environmental map holding means 33 When the environmental map holding means 33 inputs the environmental map from the environmental understanding means 31 1, if the output destination of the environmental map requires a signal in an appropriate format, the environmental map is stored in the format required by the output destination of the environmental map. Convert. Thereafter, the environmental map holding means 33 stores the environmental map for a certain period of time until the environmental map is reliably transmitted to the output destination of the environmental map.
- the visual device shown in FIG. can do.
- the output device of the environmental map can use the visual device shown in FIG. 9 as a device for specifying the positions of the moving object 2 and the stationary object 3 that can be photographed by the moving camera 10.
- the visual device shown in FIG. 10 controls the moving camera 10 to the estimated position of the object by adding the object position estimating means 34 to the visual device shown in FIG.
- the object position estimating means 3 4 inputs an environmental map from the environmental understanding means 3 1 and the recognition result is null. Select one of the environmental data, which is data, and extract the position in this environmental data. By subtracting the current position of the moving camera 10 on the environment coordinate system calculated by the camera / environmental coordinate conversion means 20 from this position, the object position estimating means 34 becomes the environment coordinate system of the moving camera 10. The relative position in the environment coordinate system of the moving object 2 and the stationary object 3 for which the recognition result has not been obtained even though the coarse edge information 1 1 2 has been generated in the past, with the current position in the above as the origin. Can be requested. However, in the case of moving object 2, moving object 2 is not always present at this position.
- this position is the estimated position of moving object 2 and stationary object 3 where moving object 2 and stationary object 3 may exist.
- a position on the appropriate environment coordinate system is generated within a range in which the mobile camera 10 can move.
- the following can be considered as a reference for generating an appropriate position.
- an arbitrary position is generated by a pseudo random number.
- Third, positions within the movable range of the moving camera 10 are sequentially generated in an appropriate order.
- the position is generated in order from the upper left position to the right, and when it reaches the right end, it descends one step and then generates the position in order to the left. After reaching the left end, go down one step and repeat to generate positions in order to the right.
- the object estimation position is output from the object position estimation means 34 to the motion control means 23.
- the motion control means 23 is changed as follows based on the motion control means 23 in the visual device according to the fourth aspect.
- the motion control means 23 includes the control command input means 24, the object position estimating means 34, the position selecting means 22 and the vibration command generating means 25. Is selected from the means having an input position in the order of. However, it is necessary for the motion control means 23 not to continuously select the input position from the object position estimation means 34.
- the visual device shown in FIG. 10 can search for the moving object 2 and the stationary object 3 existing in the range where the moving camera 10 can photograph.
- the output destination of the environmental map can use the visual device shown in FIG. 10 as a device for specifying the positions of the moving object 2 and the stationary object 3 that can be photographed by the moving camera 10.
- the visual device of FIG. 11 generates an arbitrary number of objects by adding an object counting means 35 and an object number holding means 36 to the visual device of FIG.
- the object counting means 35 When inputting an environment map from the environment understanding means 31, the object counting means 35 counts the number of environment data having a recognition result meaning a specific object in the environment map, and generates an arbitrary number of objects. Any number of recognition results in the environment data can be selected from the types of objects identified by the image recognition means 29. The number of arbitrary objects is output from the object counting means 35 to the object number holding means 36. If necessary according to the application, it is easy to change the object counting means 35 so that the type of the object to be counted can be specified from the outside.
- the object number holding means 36 inputs the arbitrary object number from the object counting means 35, the arbitrary object number is stored for a certain period of time until the arbitrary object number is reliably transmitted to the output destination of the arbitrary object number.
- the visual device shown in FIG. 11 can count the number of the specific moving objects 2 and the stationary objects 3 existing in the range where the moving camera 10 can photograph. Can be counted. As a result, as the output destination of an arbitrary number of objects, the visual device shown in FIG. 11 can be used as a device for counting the number of specific moving objects 2 and stationary objects 3 that can be photographed by the moving camera 10.
- the visual device described in claim 5 is intended to make the creation of an environmental map accurate and fast by adding a geometric analysis means 37 to the visual device of FIG. is there.
- the geometric analysis means 37 receives the formed edge information image 115 from the edge information forming means 15 and performs a geometric analysis such as a stroke extraction method, a Fourier transform, and a Hough transform to move the frame image 1. Estimate the shape of object 2 and stationary object 3 and generate geometric analysis results.
- a geometric analysis such as a stroke extraction method, a Fourier transform, and a Hough transform
- the object background separating means 16, the area normalizing means 27, and the image recognizing means 29 recognize the patterns of the moving object 2 and the stationary object 3 in the frame image-1. Therefore, the geometric analysis means 37 uses the formed edge information image 115 to determine the moving object 2 and the stationary object 3 which the object background separation means 16, the area normalizing means 27 and the image recognition means 29 are not good at.
- the image recognition means 29 can omit unnecessary pattern matching, and the environment understanding means 31 can create an environment map accurately and at high speed. And the load on the geometric analysis means 37 itself can be reduced.
- the geometric analysis result is output from the geometric analysis means 37 to the image recognition means 29 and the environment understanding means 31. Therefore, the image recognition means 29 and the environment understanding means 31 operate as follows.
- the image recognition means 29 Upon input of the geometric analysis result from the geometric analysis means 37, the image recognition means 29 first determines whether or not the image recognition means 29 is a figure to be recognized. If the result of the geometric analysis is not a target figure, the image recognition means 29 does not operate. If the geometric analysis result is the target figure, the image recognition means 29 performs pattern matching using the template image for the target figure. For example, in the case of a perceptron that learns by the error backpropagation method, the perceptron is trained for each target figure to create a learning data, and then the learning data is selected by the geometric analysis result, thereby improving the efficiency for the target figure. Pattern matching is performed.
- the environment understanding means 31 Upon input of the geometric analysis result from the geometric analysis means 37, the environment understanding means 31 first determines whether or not the image recognition means 29 is a figure to be recognized.
- the environment understanding means 31 immediately recognizes the environment data near the position on the environment coordinate system of the moving camera 10 and whose recognition result is null data. Remove from map. As a result, unnecessary environmental data is deleted from the environmental map, and the object position estimating means 34 It is not necessary to output the estimated body position. If the geometrical analysis result is the target figure, wait until the recognition result is input from the image recognition means 29.
- the visual device according to claim 5 can accurately and quickly count the number of specific moving objects 2 and stationary objects 3 existing in a range where the moving camera 10 can photograph. Can be.
- the output device of an arbitrary number of objects can use the visual device according to claim 5 as a device for counting the number of specific moving objects 2 and stationary objects 3 that can be photographed by the moving camera 10 at high speed.
- the position / size detecting means 17, the area normalizing means 27, the normalized image holding means 28 and the image recognizing means 29 are composed of an array operation unit 40 (ARRAY OPERATION UNIT). It can be implemented by using the processing device 110. Therefore, in the following, an embodiment of the data processing device 110 using the array operation unit 40 will be described, and the visual device described in claims 6 to 12 will be described with reference to the drawings.
- the array operation unit 40 generates one pixel of the output image by using one pixel of the input image and its neighboring pixels. Therefore, as shown in FIG. 13, by using the data processing device 110 in which the array operation unit 40 is arranged in a lattice in accordance with the size of the input image, the data processing device 110 Can generate an output image.
- the array operation unit 40 is abbreviated as A ⁇ U.
- the array operation unit 40 may be implemented by dedicated hardware, or may be implemented by software on a general-purpose computer. In other words, as long as the output image can be generated from the input image, the mounting means is not limited. Therefore, by indicating the algorithm of the array operation unit 40, the image processing of the data processing device 110 can be indicated.
- image storage means 12 (see FIGS. 1 and 6), image vibration means 13 (see FIGS. 3, 4 and 5), Edge information generating means 14 (see FIGS. 1 and 6) and edge information forming means 15 (see FIGS. 1 to 5 and FIGS. 7 to 12) ), Object Z background separation means 16 (see Figures 2, 4, 5, 7 and 12), position / size detection means 17 (see Figures 1 and 6), area normalization
- the mathematical expressions used in the means 27 see FIG. 7
- x, y, and w be any 2n grayscale image with width w, height h, and number of bands b, x, y, and w are the band pixel values at position P (i, j, k) ⁇ i jk , Yi jk , w; jk are represented as Expression 1, Expression 2, and Expression 3.
- Bold letters indicate vectors.
- N is a non-negative integer, w, h, b, i, j, and k are natural numbers.
- x is value at p (i, j, k), l ⁇ i ⁇ w, l ⁇ j ⁇ h, l ⁇ k ⁇ b ⁇ (1)
- y (Vijk ⁇ yijk is value at p (i, j, k) , l ⁇ i ⁇ w, l ⁇ j ⁇ h, l ⁇ k ⁇ b ⁇ (2)
- w is value at p (i, j, k), l ⁇ i ⁇ w, l ⁇ j ⁇ h, l ⁇ k ⁇ b ⁇ (3)
- the maximum value is selected from the values of each band of the pixel in the i-th row and the j-th column according to Expression 5. Since the band maximum value image is a single band image, it is handled as the image having the number of bands 1 for convenience. Therefore, the third subscript of the function B ij ⁇ ( ⁇ ) is 1.
- Equation 7 The logarithmic transformation at the position p (i, j, k) of the image x is performed according to Equation 7.
- e l is generally sufficient.
- N ij k The number of elements N ij k is always Q.
- the vibration at the position p (i, j, k) of the image X is performed in accordance with Expression 9.
- V3 ⁇ 4 fc x L x imk-N ijk x ijk (11)
- the zero point found by Expression 12 is not a place with an edge, but a place with noise, that is, a place without an edge.
- Equation 15 Assuming that the image x is an arbitrary binary image, in order to detect a pixel having a line width of 1 in the image X, calculation is performed according to Equation 15 using four neighboring pixels.
- each band pixel value of the image X can be masked using the band pixel value of the image y according to Equation 21.
- O ijk (x, y) XijkViji (21) If there are two images x and y, and images X and y are binary images, the image y can be shaped based on the image—x according to Equation 22. otherwise.
- the processing is simplified by treating the position and the movement amount of the pixel as image data. This is called position imaging.
- position imaging In the following, some functions and operations related to imaging are described.
- the operator that converts each value of m and o into band pixel values as image data is #, and the converted band pixel values are #p (1, m, o) and I do.
- the band pixel value moves from position p (i, j, k) to position p (i + 1, j + m, k + o).
- the shift amount of the band pixel value is represented as a position p (1, m, o). That is, the movement amount can be regarded as a vector from a certain position.
- # -1 # p (1, m, o) p (1, m, o).
- the movement amount p (i, j, k) can be turned 180 degrees in the opposite direction in the plane represented by the width direction and the height direction.
- T (p (i, j,)) p (-i, -j, k) (24) If there is an image x and the image x is a single-band binary image, the position p (i, j)
- the amount of movement to the position of the center of gravity in (1, 1) is calculated according to Equation 25. Note that division must be performed when calculating the center of gravity, but division is canceled when calculating the amount of movement into the vicinity of 8, so division is omitted in Equation 25.
- Giji x p ⁇ (I-i) ximi, L (m-j iml , 0) (25) ⁇ ( ⁇ , ⁇ , 1) ⁇ ⁇ ⁇ (9) p (l, m, l) ePiji (q )
- Equation 27 is used only when Equation 26 cannot be used due to image discretization.
- Equations 25, 26, and 27 the band pixel value of the displacement image of the single-band binary image X in the direction of the center of gravity can be simply described according to Equations 28 and 29. Note that the number of bands of the moving amount image is one.
- a ⁇ . 1 (x) e '(G il (x)) ,
- the band pixel value of the moving amount image in the direction opposite to the center of gravity of the image can be easily described.
- the number of bands of the moving amount image is one.
- the band pixel value of the image X is moved to the movement position indicated by the image y and then moved to the same band pixel according to Equation 31.
- the sum of the band pixel values can be converted into a grayscale image.
- Equation 32 or 33 the single band grayscale image X is moved in the direction of the center of gravity in the vicinity, and then the sum of the band pixel values moved to the same band pixel is calculated. The total can be easily described.
- image X is a binary image
- image y is a moving amount image
- the position of the destination of each band pixel value of image X can be obtained, so the destination is duplicated It is possible to find the pixel value of the band. Therefore, the band pixel value of the movable image that indicates that the destination of each band pixel value of the image X does not overlap and that there is a moving band pixel value is generated according to Expression 34. Note that the number of bands of movable images is one.
- Hiji (x, y) ⁇ for only one ⁇ p (l, m, 1) G 3 ⁇ 4i (, (34)
- Equation 36 the band pixel value of the image obtained by moving the band pixel of the image X in the direction opposite to the position of the center of gravity calculated from the binary image y is simply described according to Equation 36. can do.
- the array operation units 40 arranged in a lattice form are the same. Work in parallel in anticipation. If the array operation unit 40 arranged at the i-th row and the j-th column on the grid is AOU s j, the algorithm of AOU ij is as shown in FIG.
- step 1 201 AO U i j is arranged in the i-th row :) ′ column on the grid. This is necessary to determine the neighborhood of AO U i j, whether logical or physical.
- step 122 the neighborhood of AOU! J and initial values of variables are set.
- step 123 it is determined whether or not the digital images 111 to be sequentially input have been lost. If there is no digital image 1 1 1 (step 1 203: Y E S), the algorithm ends. If there is a digital image 1 1 1 (step 1 203: N ⁇ ), the process proceeds to step 1 204. However, when the array operation unit 40 is implemented only for a specific image size, an infinite loop may be used.
- step 124 input is waited for until digital image 111 is prepared.
- the pixels of the i-th row and the j-th column of the digital image 111 are input for the number of bands. For this reason, AOU; j requires a memory 42 for storing at least the image data for the number of bands.
- the pixel at the i-th row and the j-th column of the digital image 111 is stored so that the image can be output while waiting for input.
- step 127 the band pixel value of the digital image 111 is output. Thereafter, the flow returns to step 1203.
- the visual device corresponding to the image storage unit 12 can store the digital image 11 1 using the data processing device 110 including the array operation unit 40.
- the image vibrating means 13 (FIGS. 3, 4, and 5) according to claim 6, which is realized by the data processing device 110, is arranged in a grid pattern to vibrate the digital image 111.
- the array operation unit 40 operates synchronously and in parallel. If the array operation unit 40 arranged at the i-th row and the j-th column on the lattice is A ⁇ U i j, the algorithm of AO U i j is as shown in FIG. In step 1301, AOU j j is arranged at the i-th row and the j-th column on the grid. This is necessary to determine the neighborhood of AO U i j, whether logical or physical.
- step 1302 the neighborhood of AOU i ”and initial values of variables are set.
- step 1303 it is determined whether or not the digital images 111 to be sequentially input have been lost. If there is no digital image 1 1 1 (step 13 03: Y E S), the algorithm ends. If there is a digital image 1 1 1 (step 13 03: N ⁇ ), go to step 1 304. However, when the array operation unit 40 is implemented only for a specific image size, an infinite loop may be used.
- the pixels of the i-th row and the j-th column of the digital image 111 are inputted for the number of bands. Therefore AOU u needs a memory 4 2 for storing image data at least the number of bands.
- the pixel at the i-th row and the j-th column of the digital image 111 is moved to one of the neighboring pixels according to the function S uk (X).
- step 1306 the band pixel value of the digital image 111 is output. Thereafter, the flow returns to step 1303.
- the visual device according to claim 6 corresponding to the image vibrating means 13 vibrates the digital image 111 by using the data processing device 110 composed of the array operation unit 40. be able to.
- the edge information generating means 14 (see FIGS. 1 and 6) according to claim 7, which is realized by the data processing device 110, generates the coarse edge information image 113 from the digital image 111.
- the array operation units 40 arranged in a lattice form operate synchronously and in parallel. Assuming that the array operation unit 40 arranged at the i-th row and j-th column on the grid is AOU ij, the algorithm of AOU ij for the edge information generating means 14 is as shown in FIG.
- step 1401 A ⁇ U; j is arranged on the i-th row and the j-th column on the grid. This is necessary to determine the neighborhood of AOU i j, whether logical or physical.
- step 1402 the neighborhood of AOU i ”and initial values of variables are set.
- the neighborhood size Q used in each of the above functions may be individually determined to be 4 or 8, or the whole may be unified to 4 or 8.
- the edge information generation means 14 should deal with it by appropriately changing the neighborhood size as necessary. Can be.
- step 1403 it is determined whether the digital image 1 1 1 has been completed. If there is no digital image 1 1 1 (step 1443: Y E S), the algorithm ends. If there is a digital image 111 (step 1403: NO), the algorithm ends. However, when the array operation unit 40 is implemented for a specific number of bands and an image size, an infinite loop may be used.
- step 144 the pixels in the i-th row and the j-th column of the digital image 111 are input for the number of bands. This is because the AOU u processes collectively pixel on the column i and the row j of the digital image 1 1 1. Therefore, AOU ij requires a memory 42 for storing at least the image data of the number of bands.
- step 1405 AOU ij communicates with the nearby array operation unit 40 to perform smoothing on each band pixel value of the input digital image 111 according to the function S; " k (X). line U.
- the smoothed band pixel values are treated as band pixel values of the smoothed image.
- the function i jk (X) may be repeated several times as needed. In the case of a general color image, two times is enough.
- step 1406 logarithmic conversion is performed on each band pixel value of the smoothed image in accordance with the function Lijk (x).
- the logarithmically converted band pixel values are treated as band pixel values of the logarithmically converted image.
- step 1407 A ⁇ U U communicates with the neighboring array operation unit 40 to perform sharpening according to the function E i ′ k (X) for each band pixel value of the logarithmically transformed image. .
- the sharpened band pixel values are treated as band pixel values of the sharpened image.
- each band pixel value of one input pre-sharpened image is subtracted from each band pixel value of the sharpened image in accordance with the function Dijk (X, y).
- the band pixel value for which the difference has been calculated is treated as the band pixel value of the time difference image.
- each band pixel value of the one-input pre-sharpened image is replaced with a corresponding band pixel value of the sharpened image.
- step 1410 AOU ij communicates with the nearby array operation unit 40 to calculate Laplacian for each band pixel value of the time difference image according to the operator V jk X.
- the band pixel value for which the Laplacian has been calculated is treated as the band pixel value of the time difference Laplacian image.
- AOU ij communicates with the neighboring array operation unit 40 to extract zero points according to the function Z i ′′ k (X) for each band pixel value of the time difference Laplacian image.
- the band pixel value from which the zero point is extracted is treated as the band pixel value of the time difference zero point image.
- step 1412 the maximum value of each band pixel value is detected according to the function B i) X) for each band pixel value of the time difference Laplacian image.
- the detected maximum value band pixel value is treated as the band pixel value of the maximum value time difference zero point image.
- the number of bands is 1 for convenience.
- AOU ij communicates with the neighboring array operation unit 40 to calculate Laplacian for each band pixel value of the sharpened image according to the operator V f jk X. U.
- the band pixel value for which the Laplacian has been calculated is treated as the band pixel value of the Laplacian image.
- AOU i communicates with the neighboring array operation unit 40 to extract zero points according to the function Z i ′′ k (X) for each band pixel value of the Laplacian image.
- the band pixel value from which the zero point is extracted is treated as the band pixel value of the zero point image.
- Step 1 4 1 follow the function B ij x (x) for each band-pixel value of the Laplacian image to detect the maximum value of each band-pixel value.
- the detected maximum band pixel value is treated as the band pixel value of the maximum zero point image.
- the number of bands is 1 for convenience.
- step 1 4 for each band pixel value of the Laplacian image and each band pixel value of the time difference Laplacian image, among the band pixel values at the same position of each image according to the function M ij k (x, y) Find the maximum value.
- the detected maximum band pixel value is treated as the band pixel value of the hybrid zero point image.
- the number of bands is 1 for convenience.
- the band pixel value from which the hole has been removed is treated as the band pixel value of the hole-removed hybrid point image.
- the number of bands is 1 for convenience.
- the function F i jk (X) may be repeated several times as necessary. In the case of general force images, one time is sufficient.
- Step 1 4 1 AOU i ⁇ by communicating with the array operation units 4 0 in the vicinity of, for a band-pixel value of the hole-deleted mixed zero-point image function A; isolated points and isolated pores according jk (x) Is removed.
- the band pixel values from which isolated points and holes are removed are treated as band pixel values of the noise-removed hybrid zero-point image.
- the number of bands is 1 for convenience.
- step 1419 0 and 1 are inverted with respect to the band pixel value of the noise removal hybrid zero point image according to the function I i jk (X).
- the inverted band pixel value is treated as a band pixel value of the coarse edge information image 113.
- step 144 the band pixel value of the rough edge information image 113 is output. After that, return to step 1403.
- the visual device according to claim 7, which corresponds to the edge information generating means 14 using the data processing device 110 composed of the array operation unit 40, can convert the coarse edge information from the digital image 111. Images 1 1 3 can be generated.
- the edge information forming means 15 (see FIGS. 1 to 5, FIG. 7, and FIGS. 7 to 12) realized by the data processing device 110 9.
- a formed edge information image 1 15 composed of formed edge information 1 14 is generated from the rough edge information image 1 13 and the digital image 1 11 according to claim 8, which is composed of edge information 1 1 2 Therefore, the array operation units 40 arranged in a lattice form operate synchronously and in parallel. If the array operation unit 40 arranged at the i-th row and the j-th column on the lattice is AOU i j, the algorithm of A ⁇ U i j is as shown in FIG.
- step 1501 A ⁇ U i is arranged at the i-th row and the j-th column on the grid. This is necessary to determine the neighborhood of AO U i j, whether logical or physical.
- step 1502 the neighborhood of AO U i] and initial values of variables are set.
- the neighborhood size q used in each of the above functions may be individually determined to be 4 or 8, or the whole may be unified to 4 or 8.
- the edge information forming means 15 can change the neighborhood size as needed, if necessary, depending on the calculation time for forming the coarse edge information 112 and the number of bands of the input digital image 111. I can deal with it.
- step 1503 it is determined whether or not the digital image 111 or the coarse edge information image 113 input sequentially is lost. If there is no digital image 111 or coarse edge information image 113 (step 1503: Y E S), the algorithm ends. If any of the digital image 111 or the coarse edge information image 113 is present (step 1503: NO), the process proceeds to step 1504. However, when the array operation unit 40 is implemented for a specific number of bands and an image size, an infinite loop may be used.
- step 1504 the image of row i and column j of the digital image 1 11 and the coarse edge information image 1 1 3 Input for the number of bands.
- AOU i] collectively processes the pixels on the i-th row and the j-th column of the digital image 111 and the coarse edge information image 113.
- AOU ij requires a memory 42 for storing at least image data for the number of bands.
- step 1505 the pixel at the i-th row and the j-th column of the digital image 111 and the pixel at the i-th row and the j-th column of the coarse edge information image 113 are separated. This is because AOU i! Processes the pixel on the i-th row and j-th column of the digital image 111 and the pixel on the i-th row and the j-th column of the coarse edge information image 113 as independent image pixels. If the pixel at the i-th row and the j-th column of the digital image 111 and the pixel at the i-th row and the j-th column of the coarse edge information image 113 are separated and input from the beginning, nothing is performed.
- step 1506 AOU ij communicates with the neighboring array operation unit 40 to perform smoothing on each band pixel value of the input digital image 111 according to the function S ij k (X).
- the smoothed band pixel values are treated as band pixel values of the smoothed image.
- the function S i jk (X) may be repeated several times as needed. In the case of a general color image, two times is enough.
- step 1507 logarithmic conversion is performed on each band pixel of the smoothed image according to the function LUk (X).
- the logarithmically converted band pixel values are treated as band pixel values of the logarithmically converted image.
- step 1508 AOU ij communicates with the neighboring array operation unit 40 to sharpen each band pixel value of the logarithmically transformed image according to the function E ij k (X).
- the sharpened band pixel values are treated as band pixel values of the sharpened image.
- step 1509 AOUij communicates with the neighboring array operation unit 40 to calculate Laplacian for each band pixel value of the sharpened image according to the operator V ⁇ jkX .
- the band pixel value for which the Laplacian has been calculated is treated as the band pixel value of the Laplacian image.
- step 1510 AOUij communicates with the neighboring array operation unit 40 to form a function Z! For each band pixel value of the Laplacian image.
- the zero point is extracted according to k (X).
- the band pixel value from which the zero point is extracted is treated as the band pixel value of the zero point image.
- step 1 5 1 for each band pixel value of the zero-point image, The maximum value is detected from the area pixel values.
- the detected maximum band pixel value is treated as the band pixel value of the maximum zero point image.
- the number of bands is 1 for convenience.
- step 1512 0 and 1 are inverted with respect to the band pixel value of the maximum value zero point image according to the function IUk (X).
- the inverted band pixel value is treated as a band pixel value of the basic edge information image.
- the input band pixel value of the coarse edge information image 113 is first treated as a band pixel value of the shaped coarse edge information image, and AOUij communicates with the neighboring array operation unit 40.
- the band pixel value of the shaped rough edge information image is shaped according to the function Q i jk (X, y) using the band pixel value of the basic edge information image.
- the shaped band pixel value is treated again as the band pixel value of the shaped coarse edge information image.
- the function Q i jk (X, y) is repeated until the band pixel value of the shaped coarse edge information image no longer changes.
- the quality of the input coarse edge information image 113, and the quality required for the formed edge information image 115 it is better to terminate the calculation at an appropriate number of repetitions. good.
- step 1514 the AOU i 3 communicates with the neighboring array operation unit 40 to perform a function C i jk (x line width complementation on the band pixel value of the shaped coarse edge information image.
- the complemented band pixel values are treated as band pixel values of the formed edge information image 115.
- step 1515 the band pixel value of the formed edge information image 115 is output. Then, return to step 1503.
- the visual device according to claim 8 corresponding to the edge information forming means 15 is formed from the coarse edge information image 113 by using the data processing device 110 composed of the array operation unit 40.
- the edge information image 1 15 can be generated.
- the formation of the rough edge information image 1 13 into the formed edge information image 1 15 means that the same scene was shot from the edge information generated from the low resolution digital image 1 11 It can be regarded as estimating edge information to be generated from the high-resolution digital image 111. Therefore, for natural number ⁇ , as shown in Fig. 19, digital When the low-resolution coarse edge information image 1 17 is generated from the low-resolution digital image 1 16 in which the resolution of the image 1 1 1 is reduced to 1 / n using the edge information generation means i 4, the low-resolution coarse edge The coarse edge information image 113 can be generated by expanding the information image 117 by n times.
- the band pixel value is simply set to 0 between successive pixels in the horizontal and vertical directions. It is sufficient to fill n-1 pixels.
- the data processing device 110 that realizes the edge information forming means 15 forms the coarse edge information image 113, which is an enlargement of the low-resolution coarse edge information image 117.
- the edge information image 1 15 and a data processing device 110 for realizing the edge information forming means 15 are formed edge information images 1 1 1 which form a coarse edge information image 1 13 generated from the digital image 1 1 1. 5 is almost the same.
- the edge information forming means 15 uses the rough edge information in order to refer to which edge information of the internally generated edge information is used by the edge information forming means 15 using the digital image 11. This is because only the information image 1 1 3 is used. Therefore, when the coarse edge information image 113 obtained by enlarging the low resolution coarse edge information image 117 is input to the edge information forming means 15, the low resolution coarse edge information image 117 from the low resolution digital image 116 is obtained.
- the data processing device 110 that realizes the generated edge information generating means 13 can reduce the amount of hardware.
- the low-resolution coarse-edge information image 1 17 generated from the low-resolution digital image 1 16 with the reduced resolution of the digital image 1 1 1 1 It is possible to generate a low-resolution cut-out coarse edge information image 118 cut out around the edge information 112.
- the data processing device 110 for realizing 14 it is possible to generate the cutout formed edge information image 121.
- the data processing device 110 that implements the edge information forming means 14 can reduce the amount of hardware. As shown in FIG.
- the position / size detecting means 17 (refer to FIGS. 1 and 6) according to claim 9 realized by the data processing device 110 is provided with coarse edge information 1 1 2
- the array operation units 40 arranged in a grid are synchronized in parallel to generate the overlapping information image 13 Works. If the array operation unit 40 arranged at the i-th row and the j-th column on the lattice is AO U i], the algorithm of AOU u is as shown in FIG.
- AO U i j is arranged at the i-th row and the j-th column on the grid. This is necessary to determine the neighborhood of AO Ui, whether logical or physical.
- step 1702 the neighborhood of AOU i j and the initial values of variables are set.
- the neighborhood size Q used in each of the above functions may be determined individually, or all may be unified.
- the position / size detection means 17 may be used as necessary, depending on the calculation time for calculating the center of gravity of the object's coarse edge information 112, the size of the input coarse edge information image 113, etc. This can be dealt with by appropriately changing the neighborhood size.
- step 1703 it is determined whether or not the sequentially input coarse edge information images 113 have disappeared. If there is no coarse edge information image 1 13 (step 17 03: Y E S), the algorithm ends. If there is a coarse edge information image 113 (step 1703: NO), the flow shifts to step 1704. However, when the array operation unit 40 is implemented only for a specific image size, an infinite loop may be used.
- step 1704 the pixels in the i-th row and the j-th column of the rough edge information image 113 are input for one band. Therefore, AOU; j requires a memory 42 for storing at least one band of image data.
- step 1705 the rough edge information 1 13 of the rough edge information image 1 13 is converted into the overlap information 1 3 1 of the overlapping information image 1 32.
- Duplicate information 1 3 1 is a band pixel value corresponding to 1 or 0.
- step 1706 A ⁇ U ij communicates with the neighboring array operation unit 40 to determine the amount of movement according to the function ⁇ u i (X) for each band pixel value of the overlapping information image 132. calculate.
- the band pixel value obtained by imaging the movement amount is treated as a band pixel value of the movement amount image.
- step 1707 A ⁇ U ij communicates with the neighboring array operation unit 40 to move each band pixel value of the overlapping information image 132 in accordance with the function ⁇ u i (X).
- the shifted band pixel value is newly treated as a band pixel value of the overlapping information image 132.
- step 1708 it is determined whether or not the number of movements representing the number of repetitions from step 1705 to step 1707 has reached the specified number. If the number of times of movement has not reached the specified number of times (step 1708: N ⁇ ), the flow returns to step 1705. If the number of times of movement has reached the specified number of times (Step 1708: Y E S), the flow shifts to Step 1709.
- the number of times specified is determined by the size of the rough edge information image 113, the size of the object represented by the rough edge information 112, and the size Q of the neighborhood. If appropriate parameters are set according to the purpose of use, it does not matter if the number of specified times is overestimated, but if the number of specified times is too large, the time required to detect the position and size increases. Become.
- step 1709 the AOU ij communicates with the neighboring array operation unit 40 to calculate the movement amount for each band pixel value of the overlapping information image 132 according to the function ⁇ ′ uX ).
- the band pixel value obtained by imaging the movement amount is treated as a band pixel value of the movement amount image.
- step 1710 AOUij communicates with the neighboring array operation unit 40 to move each band pixel value of the overlapping information image 1332 in accordance with the function ⁇ 'ijj (X).
- the shifted band pixel value is newly treated as a band pixel value of the overlapping information image 132.
- step 1711 the band pixel value of the overlapping information image 1332 is output. Thereafter, the flow returns to step 1703.
- each piece of overlap information 13 1 of the overlap information image 13 2 represents the total number of coarse edge information 1 12 around the position, and as a result, the size of the object around the position Will mean.
- the visual device according to claim 9 corresponding to the placement size detecting means 17 can generate the overlapping information image 132 from the rough edge information image 113_.
- the visual device according to claim 9 can also generate the overlapping information image 132 from the formed edge information image 115 instead of the rough edge information image 113. Therefore, using the data processing device 110 composed of the array operation unit 40, the visual device corresponding to the position / size detecting means 17 is used to generate the overlapping information image 1 3 2 from the formed edge information image 1 15 Can be generated.
- the position and Z size detecting means 17 realized by the data processing device 110 can obtain the overlapping information from the object region image 144 representing the object region 141.
- a duplicate information image 1 32 representing 1 3 1 can be generated.
- each piece of overlapping information 13 1 of the overlapping information image 13 2 represents the total number of pixels of the object area 14 1 centered on that position. Means the area of the object about. Therefore, when calculating the size of the object from the overlapping information image 1 32, care must be taken, such as taking the square root of the overlapping information 13 1.
- the area normalizing means 27 (see FIG. 7) according to claim 10 realized by the data processing apparatus 110 is an object area image 1 including the object area 141.
- the array operation units 40 arranged in a lattice form operate synchronously and in parallel. . If the array operation unit 40 arranged in the i-th row and j-th column on the grid is AOU i 3, the algorithm of AOU i j is as shown in FIG.
- step 2.701 AO U ij is arranged at the i-th row and the j-th column on the grid. This is necessary to determine the neighborhood of AOU u , whether logical or physical.
- step 2702 the neighborhood of AOUU and initial values of variables are set.
- the neighborhood size q used in each of the above functions may be determined individually, or all may be unified.
- the area normalizing means 27 appropriately changes the neighborhood size as necessary. This can be dealt with.
- step 2703 it is determined whether or not the sequentially input object region image 144 or digital image 111 has disappeared. If there is no object area image 142 or digital image 111 (step 270: Y E S), the algorithm ends. If there is an object region image 142 or a digital image 111 (step 270: NO), the flow shifts to step 274. However, when the array operation unit 40 is implemented only for a specific number of bands and image sizes, an infinite loop may be used.
- step 2704 the pixel of the i-th row and the j-th column of the object area image 142 is input for one band, and the pixel of the i-th row and the j-th column of the digital image 111 is input for the number of bands.
- AOU ij collectively processes the pixel at the i-th row and the j-th column of the object area image 142 and the pixel at the i-th row and the j-th column of the digital image 111. For this reason, AOUij requires a memory 42 for storing image data of at least the total number of bands.
- step 2705 the pixel at the i-th row and the j-th column of the object area image 142 is separated from the pixel at the i-th row and the j-th column of the digital image 111.
- AOU u processes the pixel on the i-th row :) 'column of the object area image 142 and the pixel on the i-th row and j-th column of the digital image 111 as independent image pixels. If the pixel on the i-th row and the j-th column of the object area image 142 and the pixel on the i-th row and the j-th column of the digital image 111 are separated and input from the beginning, nothing is performed.
- AOUij communicates with the neighboring array operation unit 40 to calculate the movement amount for each band pixel value of the object area image 142 according to the function Rij (X).
- the band pixel value obtained by imaging the movement amount is treated as a band pixel value of the movement amount image.
- AOU; j can move according to the function Hij k (x, y) for each band pixel value of the object area image 142 by communicating with the array operation unit 40 nearby. It is possible to find the destination pixel value of the destination band. The value indicating whether the destination is a movable destination is Move It is treated as a band pixel value of a possible image.
- step 2708 by AOUi j to communicate with the array operation unit 40 in the vicinity of, the function Ui j k (x, y) for each band-pixel value of the object-area image 142 is moved to a movable target in accordance with.
- the shifted band pixel value is newly treated as a band pixel value of the object area image 142.
- step 2709 by AOUi j to communicate with the array operation unit 40 in the vicinity of, the function Ui j k (x, y) for each band-pixel value of the digital image 111 is moved to a movable target in accordance with.
- the shifted band pixel value is newly treated as a band pixel value of the digital image 111.
- step 2710 it is determined whether or not the number of movements representing the number of repetitions from step 2706 to step 2709 has reached the specified number. If the number of movements has not reached the specified number of times (step 2710: NO), the flow returns to step 2706. If the number of times of movement has reached the specified number of times (step 2710: YES), the flow shifts to step 2711. Note that the designated number is determined by the size of the digital image 111, the size of the separated object area 143 of the digital image 111, and the size Q of the neighborhood. If appropriate parameters are set according to the purpose of use, it does not matter if the number of specified times is overestimated, but if the number of specified times is too large, the time required for normalization becomes longer.
- AOUij communicates with the neighboring array operation unit 40 to calculate the average value of the neighborhood according to the function V i jk (x, y) for each band pixel value of the moved object region image 142. Interpolate. Note that both X and y become the object region image 142. The band pixel value filled with the average value is treated as a band pixel value of the normalized object region image.
- step 2712 the AOUij communicates with the neighboring array operation unit 40 to fill each band pixel value of the digital image 111 that has completed the movement with the neighboring average value according to the function Vijk (X, y).
- the separated object area 143 is converted into a normalized area 144 in the normalized image 145.
- X becomes the digital image 111
- y becomes the object region image 142.
- the band pixel value filled with the average value is treated as a band pixel value of the normalized image 145.
- step 2 7 13 it is determined whether or not the number of interpolations representing the number of repetitions from step 2 7 1 1 to step 2 7 1 2 has reached the specified number.
- step 2713 N ⁇
- step 2711 YES
- step 2714 the number of interpolations is about half of the neighborhood size Q.
- step 2714 it is determined whether or not the number of repetitions representing the number of repetitions from step 2706 to step 2713 has reached the designated number. If the number of continuations has not reached the specified number of times (step 2714: N ⁇ ), the flow returns to step 2706. If the number of continuations has reached the specified number of times (step 2714: Y E S), the process proceeds to step 2715.
- the designated number is determined by the size of the digital image 111, the size of the separated object region 144 of the digital image 111, and the size Q of the neighborhood. If appropriate parameters are set according to the purpose of use, it does not matter if the number of specified times is overestimated, but if the number of specified times is too large, the time required for normalization becomes longer.
- step 2715 the band pixel value of the normalized image 14.5 is output. Then, the process returns to step 2703.
- the visual device which corresponds to the region normalizing means 27, uses the data processing device 110 composed of the array operation unit 40, and A normalized image 144 can be generated from the digital image 111.
- the normalized image holding means 28 (see FIG. 7) realized by the data processing unit 110 stores the normalized images 145 in order to store the normalized images 145. Work synchronously and in parallel. If the array operation unit 40 arranged at the i-th row and j-th column on the grid is AO U i j, the algorithm of AO U i j is as shown in FIG.
- step 2 801 AOU u is arranged at the i-th row and the j-th column on the grid. This is necessary to determine the neighborhood of AOUij, whether logical or physical.
- step 282 the neighborhood of AOU; j and the initial values of variables are set.
- step 2803 it is determined whether or not the sequentially input normalized images 1 4 5 have been lost. You. If there is no normalized image 145 (step 2803: YES), the algorithm ends. If there is a normalized image 145 (step 2803: NO), the process proceeds to step 2804. However, when the array operation unit 40 is implemented only for a specific image size, an infinite loop may be used.
- step 2804 the pixels in the i-th row and the j-th column of the normalized image 144 are inputted for the number of bands. For this reason, AOUij requires a memory 42 for storing at least image data for the number of bands.
- step 2505 the format of the normalized image 145 is converted if the output destination device requires it. In particular, if the number of bands of the normalized image 1 4 5 is set to 1, or if the number of bands of the digital image 1 1 1 is 4 or more, the number of bands of the normalized image 1 4 5 is set to 3, making it easier to generate analog signals. Convenient in case. Otherwise do nothing.
- step 280 6 the pixels in the i-th row and the j-th column of the normalized image 145 are stored so that the image data can be reliably transmitted to the output destination device having a different processing speed.
- step 807 the band pixel value of the normalized image 145 is output. Thereafter, the flow returns to step 280 3.
- the visual device corresponding to the normalized image holding unit 28 can output the normalized image 144 using the data processing device 110 including the array operation unit 40.
- the image storage means 12 (see FIGS. 1 and 6), the image vibration means 13 (see FIGS. 3, 4 and 5), the edge information generation means 14 (see FIG. FIG. 6 and FIG. 6), edge information forming means 15 (see FIGS. 1 to 5 and FIGS. 7 to 12), object / background separating means 16 (second, fourth, fifth, and fifth). 7 to 12), position / size detecting means 17 (see FIGS. 1 and 6), area normalizing means 27 (see FIG. 7), and normalized image holding means 28 (See Fig. 7).
- the moving object 2 or the stationary object 3 enlarged over the entire normalized image 1 45 (see FIG. 24) is identified from the candidates prepared in advance, and the recognition result is obtained. Must be generated.
- the most basic method for identifying moving object 2 or stationary object 3 is to prepare as many template images 1 4 6 (see Fig. 27) of moving object 2 or stationary object 3 as By comparing 4 5 with the template image 1 4 6, we find the template image 1 4 6 that is most similar to the normalized image 1 4 5.
- the image recognizing means 29 can obtain the template image 1 4 6 most similar to the normalized image 1 45 simply by extracting arbitrary pixels from the normalized image 1 45 and the template image 1 46 and comparing them.
- the image recognition means 29 needs global processing such as the least squares method and the neural network. Since the data processing device 110 has a structure suitable for neighborhood processing, it is difficult to implement the image recognition means 29 using only the data processing device 110.
- the image recognition means 29 does not need to perform global processing throughout the entire process of generating a recognition result from the normalized image 144. That is, in the process of generating a recognition result from the result of extracting and comparing arbitrary pixels from the normalized image 144 and the template image 144, the image recognition means 29 does not need to perform global processing throughout the entire process of generating a recognition result from the normalized image 144. That is, in the process of generating a recognition result from the result of extracting and comparing arbitrary pixels from the normalized image 144 and the template image 144, the image recognition means 29 does not need to perform global processing throughout the entire process of generating a recognition result from the normalized image 144. That is, in the process of generating a recognition result from the result of extracting and comparing arbitrary pixels from the normalized image 144 and the template image 144, the image recognition means 29 does not need to perform global processing throughout the entire process of generating a recognition result from the normalized image 144. That is, in the process of generating a recognition result
- the image recognition means 29 requires global processing, but in the process of extracting and comparing arbitrary pixels from the normalized image 144 and the template image 144, the image recognition means 29 does not necessarily require global processing.
- the process of extracting and comparing arbitrary pixels from the normalized image 144 and the template image 144 is the most basic pattern matching, so if this pattern matching can be realized by neighborhood processing, this pattern Only the process of generating recognition results from matching results can be realized by a general-purpose processor that performs simple numerical calculations such as majority voting. Therefore, a method of realizing pattern matching by the data processing device 110 will be described below.
- the normalized image 1 4 5 and X First, the n-number of the template image 1 4 6 yy 2,,, and y h ,,, y n.
- the matching result image 1 4 7 Matching result (5 u compares the pixel of the i-th row and the j-th column of the normalized image 144 with the template image 144 according to Expression 37 , and has a pixel most similar to the pixel of the normalized image 144 Indicates the number of template image 1 46.
- the matching result image 1 47 is a single band image, so it is treated as an image with the number of bands 1 for convenience. It has become.
- the matching result ⁇ i j 1 generated according to Expression 37 is not necessarily unified in the entire matching result image 14 7. If there are many template images 146, the matching result image 147 is likely to be rather mosaic. Therefore, a method in which the data processing device 110 calculates a histogram for the matching result ⁇ i j i and the matching result in the vicinity of Q and converges the matching result ⁇ 5 i j will be described below.
- an arbitrary single-band image X is a matching result image 147, using a natural number g, real numbers u and V, the matching image 172 is updated according to equations 38 and 39. Since the matching result image 147 is a single band image, it is handled as an image having one band for convenience. Therefore, the third subscript of the function ⁇ U 1 (X) is 1.
- the matching result converges as follows due to the combination of the normalized image 144 and the template image 144. If about half of the pixels in the normalized image 1 4 5 are most similar to the pixels in the particular template image 1 4 6, then most of the matching results in the matching result image 1 4 7 Converge to the number 6. However, some pixel clusters of the normalized image 1 4 5 have several different template images 1
- the matching result image 1 47 will have a block of several template images 1 4 6 surrounded by 0. Further, if the normalized image 144 does not correlate with the set of template images 144, the matching result of the matching result image 144 becomes almost zero. Therefore, in the pattern matching realized by the data processing device 110, it is difficult to identify the template image 144 that is most similar to the normalized image 144, but it is difficult to identify the template image 144. It is conceivable that some similar template images 1 4 6 can be selected from. Therefore, in the process of generating the recognition result from the pattern matching result, one of the most prominent candidates is selected from the similar candidates of the template images 144 enumerated by the matching result images 147 generated by the pattern matching.
- the pattern matching according to claim 11 realized by the data processing device 110 is most similar to the normalized image 144 of the template images 144.
- the array operation units 40 arranged in a lattice form operate synchronously and in parallel. Assuming that the array operation unit 40 arranged at the i-th row and the j-th column on the lattice is AOU; j, the algorithm of A ⁇ U i is as shown in FIG.
- step 2901 AO U i j is placed on the i-th row :) 'column on the grid. This is necessary to determine the neighborhood of AOU i j, whether logical or physical.
- step 2902 the neighborhood of AOU ij and the initial values of variables are set.
- the neighborhood size Q used in each of the above functions may be determined individually, or all may be unified.
- pattern matching can be dealt with by appropriately changing the neighborhood size as necessary. .
- step 2903 it is determined whether or not the sequentially input template images 1 4 6 have disappeared. If there is no template image 1 4 6 (step 2903: Y E S), the flow shifts to step 2905. If there is a template image 146 (step 290 3: NO), the flow shifts to step 290 4.
- step 2904 the pixels of the i-th row and the j-th column of the template image 1466 are inputted for the number of bands. For this reason, A ⁇ U ij requires a memory 42 for storing at least the image data obtained by multiplying the number of bands by the number of template images 14 6. Thereafter, the flow returns to step 2903.
- step 295 it is determined whether or not the sequentially input normalized images 145 have disappeared. If there is no normalized image 1 45 (step 2905: Y E S), the algorithm ends. If there is a normalized image 145 (step 2905: NO), the flow shifts to step 290. However, when the array operation unit 40 is implemented only for a specific image size, an infinite loop may be used.
- step 2906 pixels in the i-th row and the j-th column of the normalized image 144 are inputted for the number of bands. For this reason, AOU ij requires a memory 42 for storing at least image data for the number of bands.
- step 2907 a matching result ⁇ of the matching result image 144 is calculated from the normalized image 144 and the template image 144.
- the matching result is a band pixel value indicating the number of the template image 144 closest to the normalized image 144.
- step 2908 AOUij communicates with the neighboring array operation unit 40 to update the matching result according to the function jx) for each band pixel value of the matching result image 147.
- the updated band pixel value is again treated as a band pixel value of the matching result image.
- the function ⁇ ; ”(X) indicates that the band pixel value of the matching result image 147 originally does not change. Repeat until it is no longer needed.
- the quality of the input normalized image 145, and the quality required for the updated matching result image 147 it is better to terminate the update process at an appropriate number of repetitions. good.
- step 209 the band pixel value of the matching result image 147 is output. Then, return to step 2905.
- a matching result image 1 4 7 can be generated from 5.
- a nonlinear oscillator generally causes a pull-in phenomenon.
- This pull-in phenomenon is a phenomenon in which nonlinear oscillators having different periods interact with each other and oscillate at a period of a simple constant ratio in a periodic behavior such as a limit cycle attractor.
- the vibration of one nonlinear oscillator changes, the vibration of the other nonlinear oscillator also changes, so these nonlinear oscillators are synchronized.
- the object Z background separating means 16 uses such a nonlinear oscillator pull-in phenomenon to separate the object from the background so that the edge information in the edge information image is a boundary, and the object area representing the object area Generate an image.
- van der Pol van der Pol
- the nonlinear oscillator is coupled by i j kl Te bond value calculated in accordance with Equation 41 with the nonlinear oscillators in a neighbor set ⁇ ij (Q a) contained in the vicinity of Q a. If a logarithmic table is not used, approximation by Equation 42 is also possible. ⁇ And are appropriate positive constants.
- the edge information is not sufficient to separate the object from the background, the edge information must be interpolated. For this purpose, it is necessary to find out how many non-linear oscillators are out of phase in the set of non-linear oscillators ⁇ ij (q b ) near Q b of the non-linear oscillator ⁇ u . Then, the contour parameter V ij is calculated by Equation 48.
- van 'Del' pole was described as a nonlinear oscillator. It can operate with any nonlinear oscillator that causes a pull-in phenomenon, such as a resonator.
- the parameters ⁇ ij and ⁇ i j should be replaced or added with the parameters of each nonlinear oscillator. At that time, it is only necessary to add the neighboring input sum ⁇ and the disturbance P ij to the appropriate parameters. However, in the case of a chaotic oscillator, no disturbance p ij is required.
- Equations 40 to 49 all array operation units 40 of the data processing apparatus 110 that can implement the object Z background separation means 16 (see FIGS. 2 and 7) Algorithm can be described.
- the visual device according to claim 12 corresponding to the object / background separation means 16 will be described using an algorithm of an arbitrary array operation unit 40 in the data processing device 110.
- an object inside the triangle 15 2 and the outside of the triangle 15 are obtained by using the edge information 15 1 of the triangle formed by the object background separation means 16 realized by the data processing device 110.
- the array operation units 40 arranged in a lattice form operate synchronously and in parallel. If the array operation unit 40 arranged at the i-th row and the j-th column on the grid is A ⁇ U i j, the algorithm of AOU i is as shown in FIG.
- step 1 601 AOU i j is arranged at row i and column j on the grid.
- neighbors 0) U and ckl are connected to each other by i jk i based on equations 41 and 42 .
- step 1603 set appropriate initial values for the parameters ⁇ ij and i ⁇ of the nonlinear oscillator. I do.
- step 1604 it is determined whether or not the sequentially input formed edge information image 115 has been exhausted. If there is no formed edge information image 115 (step 1604: YES), the algorithm ends. If there is the formed edge information image 115 (Step 1604: N 0), the process proceeds to Step 1605. However, when the array operation unit 40 is implemented only for a specific number of bands and an image size, an infinite loop may be used.
- step 1605 ⁇ u of the formed edge information 114 is input.
- step 1606 a disturbance p ij is calculated from 43 ij of the formed edge information 114 input immediately before in accordance with Expression 43.
- equation a to input A_ ⁇ _U kl have zeta k physician xi] k from kl array operation Interview knit 40 there is a nonlinear oscillator w kl in a neighbor set Q u (Q a), shed sum ij Calculate according to 44.
- step 1608 the parameters ⁇ ij, u of the nonlinear oscillator are calculated according to equations 45 and 46. That is, the differential equations shown in these equations are solved by the Runge-Kuy ⁇ method.
- step 1609 the output ⁇ ; j of the nonlinear oscillator is calculated in accordance with Equation 47.
- step 1610 A kl is input from AOU kl of the array operation unit 40 having the nonlinear oscillator co k i in the neighborhood set j (q b ), and the contour parameter ⁇ s 3 -is calculated according to Equation 48. calculate.
- Step 1611 the boundary parameter ⁇ i j is calculated according to Equation 49. That is, the differential equation shown in this equation is solved by the difference method or the Runge-Kutta method.
- step 1612 it is determined whether or not the number of separations representing the number of repetitions from step 1606 to step 1611 has reached the specified number. If the number of separations has not reached the specified number (step 1612: N :), the flow returns to step 1606. If the number of separations has reached the specified number (step 1612: YES), the flow shifts to step 1613.
- step 1613 the output ⁇ of the non-linear oscillator that becomes the band pixel value of the object region image 142 Output ij. Thereafter, the flow returns to step 1604.
- the following method can be used to determine the number of separations in step 1612.
- the separation is completed in a certain fixed time in almost all the formed edge information 114 regardless of the initial state of the nonlinear oscillator. It is sufficient to calculate the number of repetitions from step 1606 to step 1611. This is because if the initial state of the nonlinear oscillator is within a certain range, there is not much difference in the time until the nonlinear oscillator is synchronized by the pull-in phenomenon.
- By simply calculating the nonlinear oscillator in this way it is possible to separate the inner region 15 2 of the triangle and the outer region 15 3 of the triangle using the edge information 15 1 of the formed triangle.
- the object Z background separating means 16 separates the inside area 15 2 of the triangle and the outside area 15 3 of the triangle as shown in FIG. At this time, the phase difference between the inner region 15 2 of the triangle and the outer region 15 3 of the triangle exceeds 90 degrees and approaches 180 degrees as much as possible, so that the triangle and the background region can be separated.
- each time the formed edge information 114 is obtained the joint value is changed in a pseudo manner by the following method.
- the nonlinear oscillator c k ! A nonlinear oscillator ⁇ ; and the binding values for binding to the j i "k! (See step 1602).
- the formed edge information ⁇ u and ⁇ kl are both 1 when there is an edge and 0 when there is no edge.
- the formation edge information ⁇ from A ⁇ U k i of the array operation unit 40 to AO U ij is obtained.
- a ⁇ U ij calculates a combined value r ijkl (1-ki) and substitutes the combined value for kl (see step 1607 ).
- the boundary parameter ij acts as a scaling factor from 0 to 1 on the substituted connection value ri jk i (1 ⁇ k i) (see step 1607).
- the phase difference between the inside and outside of the edge information 154 of the dashed triangle is about 9%.
- each AOU ij calculates the output ⁇ ij of the nonlinear oscillator (see step 1609).
- the output ⁇ ij is 1, if the nonlinear oscillator of which the order is 1 among the nearby nonlinear oscillators is c k i, the parameters u and kl both become ⁇ or more.
- ⁇ u is about the same phase, and if ⁇ is a positive value, the worst case phase difference will not exceed 90 degrees.
- the maximum value of this phase difference is determined by the value of ⁇ .
- the contour parameter V ij representing the number of neighboring non-linear oscillators having approximately the same phase is calculated according to Equation 48 (see step 1610).
- the boundary parameter ij which is the magnification of the joint value, is reduced according to Equation 49, otherwise it is increased according to Equation 49. (See steps 1611). For example, in the case of around 8, if the value is between 3 and 5, the boundary parameter should be reduced according to Equation 49.
- edge information 157 of the front triangle and edge information 158 of the rear triangle are obtained.
- the phases of the nonlinear oscillators in the three regions of the inner region of the front triangle 15 9, the inner region 16 6 of the rear triangle, and the background region 16 1 of the double triangle are shifted from each other, resulting in three regions. Separated into regions.
- Fig. 33 As shown in, even if the two overlapping circular edge information 1 6 2 are dashed lines, the front circular inner area 16 3, the rear circular inner area 16 4, and the double circular background area 16 5 Separated into three.
- the visual device according to claim 12 corresponding to the object / background separating means 16 using the data processing device 110 composed of the array operation unit 40,
- the object area 14 1 and the background can be separated by using the formed edge information 114 as a boundary.
- the visual device described in claims 6 to 12 has been described.
- these visual devices can be implemented by a general-purpose computer, but when the moving object 2 is to be counted, each of the above means is executed at high speed depending on the moving speed of the moving object 2. There is a need to.
- the image storage means 12 see FIGS. 1 and 6) and the image vibration means 13 (see FIGS. 3 and 4) which process the image itself are processed.
- edge information generating means 14 (see FIGS. 1 and 6), edge information forming means 15 (see FIGS. 1 to 5 and FIGS. 7 to 12)
- Object background separation means 16 (see FIGS. 2, 4, 5, 7 and 12), position and size detection means 17 (see FIGS. 1 and 6), area normalization means 2 7 (see Fig. 7), the normalized image holding means 28 (see Fig. 7) and the image recognition means 29 (see Fig. 8) provide the image size or resolution for each of the width and height directions.
- the calculation amount increases in proportion to. Therefore, the visual device described in claims 6 to 12 may not be able to achieve the desired performance depending on the application.
- the array operation unit 40 is arranged in a grid pattern as shown in FIG. 13 in the data processing apparatus 110.
- the array operation unit 40 is further wired so as to be able to communicate with only the adjacent array operation unit 40 in the data processing device 110. That is, the four neighbors are directly wired. This makes it possible to operate at the same high speed with a smaller number of electronic components and wiring amount compared to the case of wiring between eight neighbors, and easily expandable even if the neighborhood size is expanded in the future. be able to.
- the array operation unit 40 is, as shown in FIG. 34, a processor (PROCESSOR) 41 for calculating a mathematical expression in image processing, and all parameters, constants, and functions used in the mathematical expression. And a memory for storing the operator (MEMORY) 42, and a controller 43 for communicating with a nearby array operation unit 40.
- the processor 41 is connected to an address bus 51. An arbitrary memory element and register of the memory 42 and the controller 43 can be selected by the specified address (ADDRESS).
- the processor 41 is communicably connected to the memory 42 and the controller 43 via the data bus 52 so as to be able to communicate with the memory 42 and the controller 43 in a bidirectional manner. ) Can be accessed.
- the controller 43 When the array operation unit 40 inputs a previous input data group (FRONT INPUT DATA SET) composed of one or more input pixels, the controller 43 stores the previous input data group in the memory 42. The controller 43 transmits the calculation data in the memory 42 created by the function to the adjacent array operation unit 40, and also transmits the calculation data received from the adjacent array operation unit 40 to the memory 4. 2 and, if necessary, transfer it to the array operation unit 40 other than the one input. Finally, the controller 43 outputs the image data of the output image as result data (RESULT DATA). The reason why the controller 43 is mounted on each array operation unit 40 is that the processor 41 can operate while the array operation units 40 are communicating with each other.
- a previous input data group FRONT INPUT DATA SET
- the controller 43 is capable of performing calculations at high speed and realizing high-speed processing, and because it is not necessary to change the hardware even if the number of neighboring array units 40 is changed. This is because the edge processing, that is, the exception processing for the edge pixels in the image can be automatically performed, so that the program of the processor 41 does not need to perform the edge processing, and thus becomes extremely simple.
- the address buffer (ADDRESS BUFFER) 53 is connected to the processor 41 via the address bus (ADDRESS BUS) 51.
- the address (ADDRESS) is received, and each register and other functional blocks are selected by the address decoder (ADDRESS DECODER) 54.
- the data buffer (DATA BUFFER) 55 receives data (DATA) from the processor 41 via the data bus (DATA BUS) 52, and the register selected by the address decoder 54 and the internal data bus 56. Communicate exclusively.
- the communication direction is specified by READ.
- the data is stored in the flag register 57, decoded by the flag decoder (FLAG DECODER) 58, and the adjacent array operation unit 40 as a multiple signal (SIGNALS). Sent to.
- the multiple signals are received by the FLAG ENCODER 59, analyzed and stored in the status register 60 (STATUS REGISTER), and returned to the source array operation unit 40 as reception (RECEIVE). You.
- the reception is received by the flag encoder 59 of the transmission source of the multiple signals, and as a result, completion of transmission of the multiple signals is confirmed.
- the status register 60 is selected according to the address, the contents of the status register 60 are transmitted to the processor 41 as data over the data bus 52.
- the flag encoder 59 When the flag encoder 59 receives one or more front input deliveries (FRONT INPUT SEND) corresponding to one or more input images (INPUT IMAGE), the front input data group (FRONT INPUT DATA SET) consisting of one or more input images ) Is read into the pre-input data register 61 (FRONT INPUT DATA REGISTER) prepared for the required storage capacity.
- the previous input data register 61 is selected by the address, the contents of the previous input data register 61 are transmitted to the processor 41 as data.
- the result data register (RESULT DATA REGISTER) 62 When the processor 41 completes the calculation, the result data register (RESULT DATA REGISTER) 62 is selected by the address, and the result data register 62 reads the image data of the output image as the result data (RESULT DATA).
- the flag encoder 59 sends a RESULT SEND.
- the output data register 63 (OUTPUT DATA REGISTER) is selected as an address, and the data to be transmitted to the nearby array operation unit 40 is calculated. Output data as one night (CALCURATION DATA) Evening Regis Evening Read on 6/3. Then, it is transmitted as calculation data to all adjacent array operation units 40.
- the calculation data is read into the upper input data register (UPPER INPUT DATA REGISTER) 64. Thereafter, when the upper input data register 64 is selected by the address, the contents of the upper input data register 64 are transmitted as calculation data.
- the lower input data register 65, the left input data register 66, and the right input data register 67 similarly. Operate.
- Each block of various buffers, various registers, and the address decoder 54 is a general-purpose electronic circuit.
- the flag decoder 58 and the flag encoder 59 have input / output signals as shown in FIGS. 36 and 37.
- the type (TYPE) indicates the type of the content read into the output data register 63 (OUTPUT DATA REGISTER) by 5 bits. This number of bits is a value sufficient for the array operation unit 40 to distinguish all calculation data to be transmitted and received.
- Each of the count X (COUNT-X) and the count Y (COUNT-Y) represents a 4-bit unsigned integer, and indicates the number of transfers during the array operation unit 40.
- each count becomes 0, and when the calculation data transmitted from the left and right array operation units 40 is transmitted again, the count of the flag encoder 59 is applied to each count.
- the value is obtained by adding 1 to the count Y of the flag encoder 59.
- the processor 41 specifies the direction of transmission of the output data register 63 in the upper, lower, left and right directions to the transmission flag (SEND FLAG) of the flag register 57
- the output data register 63 is set to
- the flag decoder 58 outputs a send (SEND) in accordance with the direction specified by the send flag.
- the delivery flag is represented by 4 bits, and when the calculation data of the array operation unit 40 is transmitted to the array operation unit 40 on all sides, the processor 41 sets 1 1 1 1 and the array operation unit 40 on the right side To transfer the transmitted calculation data to the upper and lower left sides, set processor 4 1 to 1 1 1 0.To transfer the calculation data from the left to upper and lower right sides, set 1 1 0 1, To transfer from the lower side to the upper side, set to 10000. To transfer from the upper side to the lower side, set to 0100. This not only eliminates duplication in the transfer, but also allows efficient transfer, and clarifies the rules for determining the transfer direction. By combining the type, count—X and count—Y, the flag encoder Can determine which type of calculation data has been transmitted from which array calculation unit 40.
- the flag decoder 58 receives the result decoding (RESULT DECODING) and transmits the result delivery (RESULT SEND) at the same time that the calculation data is read into the result data register 62 as the result decoding.
- the flag encoder 59 When the flag encoder 59 receives the delivery in any one of the four directions, the flag encoder 59 receives the type of the receiving direction, the count X, and the count Y, and updates the content of the status register 60 in that portion. At the same time as this update, set the reception to 1 in the receiving direction and send.
- the flag encoder 59 of the array operation unit 40 of the transmission source receives the data at the moment when the reception becomes 1, and updates the reception status (RECEIVE STATUS) of the status register 60.
- the processor 41 can determine which input data register has a valid calculation data stored therein only by checking the reception status of the status register 60. .
- the processor 41 can read the data from the upper input data register 64 by specifying the address, but at the same time, the address decoder 5
- the upper decoding (UPPER DECODING) is sent from 4 to the flag encoder 59, the upper part of the reception status is returned to 0, and the reception toward the upper side is transmitted as 0. The same applies to the lower left and right sides. If at least one of the flag encoders 59 receives the previous input transmission for the input image, the status register 60 (FRONT INPUT SEND STATUS) for the input image corresponding to the received previous input transmission To 1.
- the address decoder 54 sends the pre-decoding (FRONT DECODING) to the flag encoder 59, and the received previous input is transmitted. Set the previous input delivery status corresponding to to 0.
- the processor 41 reads the contents of the status register 60 so that the latest input image is stored in the previous input register 61. Can be determined whether or not.
- FIG. 38 shows an algorithm when the processor 41 transmits calculation data to the array operation units 40 on the four sides via the controller 43.
- FIG. 38 shows processing by program control by the processor 41 and mixing with hardware logic by the flag decoder 58 and the flag encoder 59.
- the processor 41 reads the contents of the status register 60.
- the processor 41 determines the type, count, and transmission direction of the data to be transmitted to the adjacent array operation unit 40, and writes the content to the flag register 57.
- step 74 the processor 41 writes data to be transmitted to the adjacent array operation unit 40 to the output data register 63.
- step 75 the contents of the output data register 63 are transmitted to the adjacent array operation unit 40 as calculation data.
- step 76 the transmission is set to 1 only in the direction specified by the transmission flag of the flag register 57 and transmitted.
- the one-time transmission algorithm of the processor 41 ends. The processor 41 starts this transmission algorithm each time the data to be transmitted is updated in the memory 42.
- FIG. 39 shows an algorithm when the controller 43 receives the calculation data from the upper array operation unit 40.
- FIG. 39 shows processing by the hardware logic by the flag decoder 58 and the flag encoder 59.
- the flag encoder 59 inputs the delivery.
- the flag encoder 59 determines whether the delivery is 1 or not. If NO, the process ends. If YES, proceed to step 83.
- the upper input register 64 reads the calculation data transmitted from the upper side.
- the flag encoder 59 sets the reception status for the upper side of the status register 60 to 1 and simultaneously sets the reception to 1 and transmits it to the upper array operation unit 40. The same applies to the lower left and right sides.
- Controller 43 is always an array operation unit Monitor the delivery from the server 40 and start the receiving algorithm each time the delivery is received.
- the algorithm when the processor 41 receives the data from the upper input data register 64 is shown in FIG. FIG. 40 shows processing by program control by the processor 41 and mixing with hardware logic by the flag decoder 58 and the flag encoder 59.
- the processor 41 reads the contents of the status register 60.
- step 92 it is determined whether or not the reception status for the upper side of the read contents is “1”. If NO, the process ends. If YES, proceed to step 93.
- step 93 the processor 41 reads the data from the upper input data register 64.
- step 94 the flag encoder 59 sets the reception status for the upper side of the status register 60 to 0, and simultaneously sets the reception to 0 and transmits it to the upper array operation unit 40.
- one reception algorithm of the processor 41 ends.
- the processor 41 monitors the contents of the status register 60 at regular intervals, and starts this receiving algorithm each time the reception status of either the upper, lower, left or right is 1. Even if the processor 41 does not monitor the contents of the status register 60 at regular intervals, it can be implemented by interrupt processing.
- This array operation unit 40 has been described mainly on the assumption that one output image is generated from one or more input images.However, depending on the application, a circuit is provided so as to be able to output calculation data during the calculation, depending on the application. Need to change. In this case, the result delivery of the flag decoder 58 is increased by the number of calculation data to be output, and only the result delivery corresponding to the calculation data read into the result data register 62 is set to 1. All you have to do is change the program.
- the amoeba when it is difficult to cut out a single moving object or a part of the moving object from a moving image as a lump area using lightness, saturation, hue, etc. Can count the number of the moving objects. For example, when counting living transparent amoeba, it is not possible to color the entire amoeba, of course. Also, even if the light source and the background are appropriately colored, the amoeba will be the same color, or the center and the edge of the amoeba will have different colors due to light refraction, reflection, etc., to obtain an image where only the ameba is filled It is also difficult.
- the present invention can also be used for a moving object that can be easily distinguished from the background by color information such as a tadpole human face. Since the number of moving objects can be counted without particularly changing the moving object, the light source, and the background, the moving object can be photographed.
- the object counting device can be realized at low cost.
- it can be used for pre-processing whenever it is determined whether or not there is a moving object in a moving image.
- it can also be used for preprocessing when recognizing a moving object, thereby realizing an object recognition device at low cost. You can also.
- the invention when it is difficult to cut out a single object or a part thereof from a still image or a frame image of a moving image as a lump area using lightness, saturation, hue, and the like,
- the invention can count the number of said objects. For example, when counting transparent beads, it is not possible to color the beads, of course. Even if the light source and background are set to appropriate colors, the beads will be the same color, or the center and edge of the beads will be different colors due to light refraction and reflection, etc. It is also difficult to obtain. In most cases, refraction or reflection of light causes a unique brightness value to appear at the outline of the bead.
- the present invention If used, the entire bead can be cut out from the background by generating edge information from this unique luminance value, so that the number of beads can be easily counted.
- microorganisms such as daphnia and cells such as leukocytes and sperm.
- the present invention can also be applied to an object that can be easily distinguished from the background by color information such as a face of a tadpole. In this way, the number of objects can be counted without particularly modifying the objects, the light source, and the background.
- the present invention By connecting the present invention to an existing device capable of photographing the object, the object counting for the object can be performed. The device can be realized at low cost.
- the present invention can be used for preprocessing whenever it is determined whether or not there is an object in a still image.
- it can also be used for preprocessing when recognizing the object, realizing the object recognition device at low cost. You can also.
- it is difficult to cut out a moving object and a stationary object alone or a part thereof from a moving image as a lump area using lightness, saturation, hue, and the like.
- the present invention can count the number of moving objects and the number of all objects among the objects.
- the present invention when counting the number of living amoebae and all amoebae among the transparent amoebae, it is assumed that the amoebae that have moved during a certain period of time are alive.
- the present invention reduces the number of moving amoeba and the number of all ameba Can be counted.
- microorganisms such as daphnia and cells such as leukocytes and sperm.
- the present invention can also be used for objects that can be easily distinguished from the background by color information such as a human face.
- the number of moving objects and the number of all objects can be counted by one device, so that an object counting device for counting the number of moving objects and the number of all objects can be realized at low cost.
- it can also be used for preprocessing such as determining whether a moving object or a stationary object is present in a moving image.
- preprocessing such as determining whether a moving object or a stationary object is present in a moving image.
- by directly extracting the object region separated from the background from the present invention and inputting it to another device it can also be used for preprocessing when recognizing a moving object or a stationary object. It can also be realized.
- the present invention counts the number of moving objects and the number of all objects among the objects, and And the number or ratio of stationary objects can be calculated. For example, when determining the survival rate of a transparent amoeba, it is assumed that the amoeba that has moved for a certain period of time is considered to be alive.
- the present invention counts the number of moving amoebae and the number of all amoebae by using two frame images separated by a certain time or two still images taken at a certain time interval in a moving image So that the survival rate of amoeba can be easily obtained.
- microorganisms such as daphnia and cells such as leukocytes and sperm.
- the present invention can also be used for objects that can be easily distinguished from the background by color information such as tadpoles and human faces. In this way, the number of moving objects and the number of stationary objects can be counted by one device, so that a device for calculating the ratio of moving objects to stationary objects can be realized at low cost.
- it can also be used for preprocessing whenever it is determined whether a moving object or a stationary object is present in a moving image. Note that by directly extracting the object region separated from the background from the present invention and inputting it to another device, it can also be used for preprocessing when recognizing a moving object or a stationary object. It can be realized at low cost.
- the present invention relates to a mobile camera using a moving image composed of an arbitrary band among three primary color wavelengths, visible light wavelengths, infrared wavelengths, ultraviolet wavelengths, and all other electromagnetic waves.
- the mobile camera can be controlled externally, it can be used in combination with a computer. Therefore, the present invention can be used as follows. For example, when monitoring objects such as people, cars and luggage in places with a lot of noise such as entrances and outdoors, in addition to places with stable lighting such as corridors, the present invention uses these objects at an appropriate magnification. You can control the moving camera as you can. In addition, the present invention is useful for preventing shoplifting by intensively photographing customers in places with little movement such as convenience stores and supermarkets. Since monitoring of an object over a wide range can be performed by a single device, the object monitoring device can be realized at low cost. In addition, it can be used for preprocessing whenever it is determined whether or not a specific object is present in a moving image.
- the image can be used for preprocessing when recognizing a moving object or a stationary object. It is possible to implement an object recognition device at low cost.
- the present invention provides a mobile camera using a moving image composed of an arbitrary band among three primary color wavelengths, visible light wavelengths, infrared wavelengths, ultraviolet wavelengths, and all other electromagnetic waves.
- An object in a photographable range can be searched. These objects are classified into several types by appropriate recognition methods, and are recorded together with attributes such as position and time. Therefore, it is possible to count the number of specific objects within a range where the moving power camera can photograph within a certain period of time, or to record the trajectory of the movement of the objects. For example, in a karaoke box or a shopping street, the present invention can measure the number of customers and the traffic volume.
- an industrial mouth pot can search for a workpiece in a production line, stop an operation by detecting an approaching human, and the like, and achieve a humanoid shape.
- the visual function of a mobile robot such as a robot or a guide dog robot can be realized.
- the present invention can detect an object located in a blind spot of a driver and issue an alarm, and can also detect a moving object such as an approaching car. It can detect in advance, or record when a traffic accident occurs. People in wheelchairs can widen their sights without turning their neck, so they can detect in advance bicycles and cars approaching from behind, and prevent accidents. It is also useful for stopping. Further, by installing the present invention at the entrance of a train or the like, the present invention can also find a passenger who is sandwiched between doors or near a door.
- the present invention can also find an object that has fallen on the track or warn a person on the side of the track on the platform. Also, if the present invention is installed in a rescue mouth pot, the rescue mouth pot finds a person who has been distressed on the coast, at sea, and in a river, etc., and automatically approaches the person to carry a life jacket or rope to the victim, The victim can be transported to a safe place. This rescue mouth pot can also be used when searching for people under the rubble under earthquakes.
- the object search, the counting, and the recording of the movement trajectory can be performed by one device, so that the object search device, the object recognition device, the object counting device, the object recording device, and the like can be realized at low cost.
- the present invention inputs each pixel of the digital image in parallel, vibrates the digital image in image units or pixel units, and then outputs each pixel of the digital image in parallel.
- a digital image can be vibrated at a high speed.
- the present invention can achieve the required processing speed.
- the coarse edge information is output in parallel. it can.
- the generated rough edge information is not always accurate because it is affected by the moving direction and moving speed of the object and the difference between the color information of the object and the background, but it is not always accurate even in a noisy environment such as outdoors under sunlight. Since the rough edge information of a moving object having an arbitrary shape can be generated without correcting the image, the present invention can be used without limiting the use environment.
- the present invention can treat infrared rays, ultraviolet rays, and even radiations in the same manner as visible light wavelengths, particularly, three primary color wavelengths, and the effect of noise can be reduced by increasing the number of bands.
- the present invention can provide the contour, position, and size of a moving object at high speed and at low cost.
- the present invention works effectively as preprocessing for a visual recognition device that does not limit the object.
- the present invention inputs the coarse edge information and each pixel of the digital image in parallel, generates the formed edge information from the coarse edge information using the digital image, and then forms the image. Edge information can be output in parallel.
- the rough edge information generated by an arbitrary method can be formed into clearer and more accurate edge information, so that the load required for the means for generating the rough edge information can be easily reduced.
- the present invention is not so affected by the quality of the coarse edge information, the coarse edge information generated from a specific area of the low resolution digital image obtained by reducing the resolution of the digital image can be formed into clearer and more accurate edge information. Wear.
- the present invention can generate high-definition edge information for an object in a low-magnification digital image without increasing the amount of hardware and the amount of calculation.
- an object recognition device that has been using a wide-angle camera and a high-definition camera can realize object recognition with one camera.
- the present invention inputs edge information in parallel, detects the position and size of the object represented by the edge information, and then parallelizes the position and size of the object as overlapping information. Can be output to With the same or better quality as the position and size of multiple objects in an image detected with a large amount of hardware and computation, the present invention uses the position and size from the edge information of multiple objects. Can be detected all at once. It is also used for preprocessing to detect the position and size of multiple objects in a video from a frame image of a moving image captured by a video camera or a still image captured by a digital camera. An image pattern recognition algorithm can be realized at high speed and at low cost.
- the position and size of the object represented by each pixel of the object area image can be output in parallel as duplicate information.
- Akira can simultaneously detect the position and size from a plurality of object areas.
- a pattern recognition algorithm for a still image and a moving image can be realized at high speed and at low cost.
- each pixel of the object region image and each pixel of the digital image are input in parallel, and after normalizing the object region in the digital image, each of the normalized image Pixels can be output in parallel.
- the present invention masks using a region other than the object region of the object region image with a quality comparable to or higher than that of a normalized image having a large amount of hardware and a large amount of calculation, which is vulnerable to misregistration generated with a large amount of hardware.
- the object area of the resulting digital image can be normalized while filling gaps according to the size of the digital image. It is also used for preprocessing for normalizing a frame object of a moving image captured by a video camera or a specific object cut out from a still image captured by a digital camera. Algorithms can be realized at high speed and at low cost.
- the present invention inputs each pixel of several template images in parallel, inputs each pixel of the normalized image in parallel, and matches the template image with the pattern matching. After that, each pixel of the matching result image can be output in parallel. If the reproducibility of the normalized image is high for the same object having different positions and sizes, the present invention can select some template images similar to the normalized image only by neighborhood processing. Therefore, the present invention can minimize global processing such as least square error and neural network, and can realize a high-speed and low-cost pattern recognition algorithm for still images and moving images.
- the present invention inputs the formed edge information in parallel, separates the object region and the background region using the nonlinear oscillator, and then parallelizes each pixel of the object region image. Can be output.
- the present invention does not require any preprocessing other than generation of edge information in advance for a digital image, does not depend on the shape, position, and orientation of an object in a digital image, It is possible to separate the object area from the background area regardless of whether or not they are dashed or intersecting, and even if the object areas in the digital image overlap.
- the present invention also makes it easy to implement hardware using digital technology, and enables high-speed processing suitable for real-time image processing.
- the present invention can operate the visual device at high speed. In particular, even if the image size increases, the processing time does not change due to parallelism. Thus, visual devices can be used for applications that require real-time performance.
- the plurality of means described in claims 6 to 12 can be realized by the same chip only by changing the program, the present invention can manufacture a visual device at low cost.
- a plurality of means described in claims 6 to 12 can be realized by one chip, this chip can be easily incorporated in a mobile camera, and the present invention can increase convenience. it can.
- signals are input and output in pixel units, the amount of wiring can be reduced by stacking a plurality of chips. Therefore, the present invention can easily improve the processing performance by technical innovations such as three-dimensional VLSI.
Description
Claims
Priority Applications (6)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
EP99943301A EP1113388B1 (en) | 1998-09-10 | 1999-09-10 | Visual device |
AU56500/99A AU763178B2 (en) | 1998-09-10 | 1999-09-10 | Visual device |
IL14158999A IL141589A (en) | 1998-09-10 | 1999-09-10 | Optical counting device using edge analysis |
JP2000570722A JP4324327B2 (ja) | 1998-09-10 | 1999-09-10 | 視覚装置 |
DE69935437T DE69935437T2 (de) | 1998-09-10 | 1999-09-10 | Visuelle vorrichtung |
US09/786,820 US6856696B1 (en) | 1998-09-10 | 1999-09-10 | Visual device |
Applications Claiming Priority (12)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP10/257327 | 1998-09-10 | ||
JP25732798 | 1998-09-10 | ||
JP11/145638 | 1999-05-25 | ||
JP14563899 | 1999-05-25 | ||
JP20973899 | 1999-07-23 | ||
JP11/209738 | 1999-07-23 | ||
JP25098699 | 1999-09-06 | ||
JP11/250986 | 1999-09-06 | ||
JP11/250990 | 1999-09-06 | ||
JP25099099 | 1999-09-06 | ||
JP11/253634 | 1999-09-07 | ||
JP25363499 | 1999-09-07 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2000016259A1 true WO2000016259A1 (fr) | 2000-03-23 |
Family
ID=27553000
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP1999/004975 WO2000016259A1 (fr) | 1998-09-10 | 1999-09-10 | Dispositif visuel |
Country Status (9)
Country | Link |
---|---|
US (1) | US6856696B1 (ja) |
EP (1) | EP1113388B1 (ja) |
JP (1) | JP4324327B2 (ja) |
KR (1) | KR100402361B1 (ja) |
CN (1) | CN1292383C (ja) |
AU (1) | AU763178B2 (ja) |
DE (1) | DE69935437T2 (ja) |
IL (1) | IL141589A (ja) |
WO (1) | WO2000016259A1 (ja) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2002073538A1 (fr) * | 2001-03-13 | 2002-09-19 | Ecchandes Inc. | Dispositif visuel, compteur asservi et capteur d'images |
CN100338598C (zh) * | 2001-03-13 | 2007-09-19 | 伊强德斯股份有限公司 | 视觉装置、联动式计数器及图象检测器 |
CN107040689A (zh) * | 2015-12-01 | 2017-08-11 | 夏普株式会社 | 图像读取装置 |
CN109598328A (zh) * | 2018-11-21 | 2019-04-09 | 山东农业大学 | 一种散布式千粒计数方法、系统、装置及终端 |
Families Citing this family (34)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB0118020D0 (en) * | 2001-07-24 | 2001-09-19 | Memco Ltd | Door or access control system |
US7028175B2 (en) * | 2003-04-09 | 2006-04-11 | Microsoft Corporation | System and method for computer hardware identification |
US7756322B2 (en) * | 2003-08-18 | 2010-07-13 | Honda Motor Co., Ltd. | Picture taking mobile robot |
US7756301B2 (en) * | 2005-01-26 | 2010-07-13 | Honeywell International Inc. | Iris recognition system and method |
JP4428067B2 (ja) * | 2004-01-28 | 2010-03-10 | ソニー株式会社 | 画像照合装置、プログラム、および画像照合方法 |
JP2006098803A (ja) * | 2004-09-29 | 2006-04-13 | Toshiba Corp | 動画処理方法、動画処理装置および動画処理プログラム |
US7359552B2 (en) * | 2004-12-15 | 2008-04-15 | Mitsubishi Electric Research Laboratories, Inc. | Foreground detection using intrinsic images |
US7796780B2 (en) * | 2005-06-24 | 2010-09-14 | Objectvideo, Inc. | Target detection and tracking from overhead video streams |
US7801330B2 (en) * | 2005-06-24 | 2010-09-21 | Objectvideo, Inc. | Target detection and tracking from video streams |
JP4696991B2 (ja) * | 2006-03-22 | 2011-06-08 | 日産自動車株式会社 | 動き検出方法および動き検出装置 |
NL2002462C2 (nl) * | 2009-01-29 | 2010-07-30 | Nedap Nv | Werkwijze voor het detecteren, in het bijzonder tellen van dieren. |
US8150902B2 (en) * | 2009-06-19 | 2012-04-03 | Singular Computing Llc | Processing with compact arithmetic processing element |
KR101105034B1 (ko) * | 2010-02-09 | 2012-01-16 | 주식회사 팬택 | 촬영 기능을 보유한 단말기 |
JP5321497B2 (ja) * | 2010-02-22 | 2013-10-23 | 株式会社デンソー | 白線認識装置 |
EP2579569B1 (en) | 2010-06-04 | 2017-08-09 | Panasonic Intellectual Property Corporation of America | Image processing device, image processing method, integrated circuit, and program |
US8730396B2 (en) * | 2010-06-23 | 2014-05-20 | MindTree Limited | Capturing events of interest by spatio-temporal video analysis |
US9218316B2 (en) | 2011-01-05 | 2015-12-22 | Sphero, Inc. | Remotely controlling a self-propelled device in a virtualized environment |
US10281915B2 (en) | 2011-01-05 | 2019-05-07 | Sphero, Inc. | Multi-purposed self-propelled device |
US9429940B2 (en) | 2011-01-05 | 2016-08-30 | Sphero, Inc. | Self propelled device with magnetic coupling |
US9090214B2 (en) | 2011-01-05 | 2015-07-28 | Orbotix, Inc. | Magnetically coupled accessory for a self-propelled device |
US8751063B2 (en) | 2011-01-05 | 2014-06-10 | Orbotix, Inc. | Orienting a user interface of a controller for operating a self-propelled device |
KR20150012274A (ko) | 2012-05-14 | 2015-02-03 | 오보틱스, 아이엔씨. | 이미지 내 원형 객체 검출에 의한 계산장치 동작 |
US9827487B2 (en) | 2012-05-14 | 2017-11-28 | Sphero, Inc. | Interactive augmented reality using a self-propelled device |
US9292758B2 (en) | 2012-05-14 | 2016-03-22 | Sphero, Inc. | Augmentation of elements in data content |
US10056791B2 (en) | 2012-07-13 | 2018-08-21 | Sphero, Inc. | Self-optimizing power transfer |
TW201425871A (zh) * | 2012-12-21 | 2014-07-01 | Zong Jing Investment Inc | 測距方法及電腦程式產品 |
WO2015017304A1 (en) * | 2013-07-30 | 2015-02-05 | Kodak Alaris Inc. | System and method for creating navigable views of ordered images |
US9829882B2 (en) | 2013-12-20 | 2017-11-28 | Sphero, Inc. | Self-propelled device with center of mass drive system |
CN106131537B (zh) * | 2016-08-17 | 2018-11-20 | 重庆转购科技有限公司 | 一种3d智能拍摄系统和方法 |
US10504007B2 (en) * | 2017-10-27 | 2019-12-10 | Facebook, Inc. | Determination of population density using convoluted neural networks |
CN107895344B (zh) * | 2017-10-31 | 2021-05-11 | 深圳市森国科科技股份有限公司 | 视频拼接装置及方法 |
JP7089179B2 (ja) | 2018-08-30 | 2022-06-22 | 富士通株式会社 | 画像認識装置、画像認識方法および画像認識プログラム |
KR102426726B1 (ko) * | 2019-11-19 | 2022-07-29 | 한국기계연구원 | 비접촉식 위치 정보 획득 방법 및 장치 |
CN112200027B (zh) * | 2020-09-27 | 2022-07-15 | 卡斯柯信号有限公司 | 一种基于机器视觉的自身移动状态识别方法 |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS61206079A (ja) * | 1985-03-09 | 1986-09-12 | Fujitsu Ltd | 変形画像表示方式 |
JPS6482184A (en) * | 1987-09-24 | 1989-03-28 | Japan Res Dev Corp | Pattern recognizing device |
JPH04128604A (ja) * | 1990-09-19 | 1992-04-30 | Ricoh Co Ltd | エッジ抽出装置 |
JPH05252437A (ja) * | 1991-07-24 | 1993-09-28 | C S K Sogo Kenkyusho:Kk | 画像処理方法および装置 |
JPH05324954A (ja) * | 1991-11-28 | 1993-12-10 | Nec Corp | 駐車台数計数装置 |
JPH07146937A (ja) * | 1993-11-25 | 1995-06-06 | Matsushita Electric Works Ltd | パターンマッチング方法 |
JPH07175934A (ja) * | 1993-12-17 | 1995-07-14 | Tokyo Gas Co Ltd | 流体の画像処理解析装置 |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US5163133A (en) * | 1987-02-17 | 1992-11-10 | Sam Technology, Inc. | Parallel processing system having a broadcast, result, and instruction bus for transmitting, receiving and controlling the computation of data |
JP3038051B2 (ja) * | 1991-06-28 | 2000-05-08 | 日本放送協会 | 動画像領域抽出装置 |
US5519793A (en) * | 1992-11-05 | 1996-05-21 | The United States Of America As Represented By The Secretary Of The Interior | Apparatus and method for computer vision measurements |
US5687249A (en) * | 1993-09-06 | 1997-11-11 | Nippon Telephone And Telegraph | Method and apparatus for extracting features of moving objects |
US6400831B2 (en) * | 1998-04-02 | 2002-06-04 | Microsoft Corporation | Semantic video object segmentation and tracking |
-
1999
- 1999-09-10 JP JP2000570722A patent/JP4324327B2/ja not_active Expired - Fee Related
- 1999-09-10 DE DE69935437T patent/DE69935437T2/de not_active Expired - Lifetime
- 1999-09-10 WO PCT/JP1999/004975 patent/WO2000016259A1/ja active IP Right Grant
- 1999-09-10 US US09/786,820 patent/US6856696B1/en not_active Expired - Fee Related
- 1999-09-10 IL IL14158999A patent/IL141589A/xx not_active IP Right Cessation
- 1999-09-10 EP EP99943301A patent/EP1113388B1/en not_active Expired - Lifetime
- 1999-09-10 AU AU56500/99A patent/AU763178B2/en not_active Ceased
- 1999-09-10 KR KR10-2001-7003103A patent/KR100402361B1/ko not_active IP Right Cessation
- 1999-09-10 CN CNB998107972A patent/CN1292383C/zh not_active Expired - Fee Related
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS61206079A (ja) * | 1985-03-09 | 1986-09-12 | Fujitsu Ltd | 変形画像表示方式 |
JPS6482184A (en) * | 1987-09-24 | 1989-03-28 | Japan Res Dev Corp | Pattern recognizing device |
JPH04128604A (ja) * | 1990-09-19 | 1992-04-30 | Ricoh Co Ltd | エッジ抽出装置 |
JPH05252437A (ja) * | 1991-07-24 | 1993-09-28 | C S K Sogo Kenkyusho:Kk | 画像処理方法および装置 |
JPH05324954A (ja) * | 1991-11-28 | 1993-12-10 | Nec Corp | 駐車台数計数装置 |
JPH07146937A (ja) * | 1993-11-25 | 1995-06-06 | Matsushita Electric Works Ltd | パターンマッチング方法 |
JPH07175934A (ja) * | 1993-12-17 | 1995-07-14 | Tokyo Gas Co Ltd | 流体の画像処理解析装置 |
Non-Patent Citations (1)
Title |
---|
See also references of EP1113388A4 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2002073538A1 (fr) * | 2001-03-13 | 2002-09-19 | Ecchandes Inc. | Dispositif visuel, compteur asservi et capteur d'images |
EP1378862A1 (en) * | 2001-03-13 | 2004-01-07 | Ecchandes Inc. | Visual device, interlocking counter, and image sensor |
CN1301491C (zh) * | 2001-03-13 | 2007-02-21 | 伊强德斯股份有限公司 | 视觉装置、联动式计数器及图象检测器 |
CN100338598C (zh) * | 2001-03-13 | 2007-09-19 | 伊强德斯股份有限公司 | 视觉装置、联动式计数器及图象检测器 |
CN100385460C (zh) * | 2001-03-13 | 2008-04-30 | 伊强德斯股份有限公司 | 联动式计数器及联动装置 |
JP2008243233A (ja) * | 2001-03-13 | 2008-10-09 | Ecchandesu:Kk | 視覚装置 |
EP1378862A4 (en) * | 2001-03-13 | 2009-06-17 | Ecchandes Inc | VISUAL DEVICE, SECURED COUNTER AND IMAGE SENSOR |
US7664220B2 (en) | 2001-03-13 | 2010-02-16 | Ecchandes Inc. | Visual device, interlocking counter, and image sensor |
JP4625513B2 (ja) * | 2001-03-13 | 2011-02-02 | 株式会社エッチャンデス | 視覚装置 |
CN107040689A (zh) * | 2015-12-01 | 2017-08-11 | 夏普株式会社 | 图像读取装置 |
CN109598328A (zh) * | 2018-11-21 | 2019-04-09 | 山东农业大学 | 一种散布式千粒计数方法、系统、装置及终端 |
CN109598328B (zh) * | 2018-11-21 | 2023-09-12 | 山东农业大学 | 一种散布式千粒计数方法、系统、装置及终端 |
Also Published As
Publication number | Publication date |
---|---|
KR100402361B1 (ko) | 2003-10-17 |
AU763178B2 (en) | 2003-07-17 |
DE69935437D1 (de) | 2007-04-19 |
IL141589A0 (en) | 2002-03-10 |
IL141589A (en) | 2005-11-20 |
US6856696B1 (en) | 2005-02-15 |
EP1113388B1 (en) | 2007-03-07 |
CN1292383C (zh) | 2006-12-27 |
EP1113388A4 (en) | 2005-04-06 |
KR20010085779A (ko) | 2001-09-07 |
DE69935437T2 (de) | 2007-11-15 |
EP1113388A1 (en) | 2001-07-04 |
AU5650099A (en) | 2000-04-03 |
CN1317124A (zh) | 2001-10-10 |
JP4324327B2 (ja) | 2009-09-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP4324327B2 (ja) | 視覚装置 | |
CN109145798B (zh) | 一种驾驶场景目标识别与可行驶区域分割集成方法 | |
WO2020119661A1 (zh) | 一种目标检测方法、装置以及行人检测方法、系统 | |
EP3499414B1 (en) | Lightweight 3d vision camera with intelligent segmentation engine for machine vision and auto identification | |
KR101191308B1 (ko) | 지능형 운송 시스템의 도로 및 차선 검출 시스템 및 방법 | |
Fabian | An algorithm for parking lot occupation detection | |
CN111539983A (zh) | 基于深度图像的运动物体分割方法及系统 | |
Othmane | Traffic surveillance system for vehicle detection using discrete wavelet transform | |
Pawar et al. | Morphology based moving vehicle detection | |
Bhardwaj et al. | Traffic control system for Smart City using image processing | |
Wasala et al. | Real-time HOG+ SVM based object detection using SoC FPGA for a UHD video stream | |
CN115147450B (zh) | 基于运动帧差图像的移动目标检测方法及检测装置 | |
CN115588187A (zh) | 基于三维点云的行人检测方法、装置、设备以及存储介质 | |
JP4201958B2 (ja) | 動画像のオブジェクト抽出装置 | |
JP2001148023A (ja) | 視覚装置及び視覚方法 | |
Quiros et al. | Localization of license plates using optimized edge and contour detection technique | |
JP2001148022A (ja) | 視覚装置 | |
JP2001256477A (ja) | 情報収集装置 | |
Kallasi et al. | Object detection and pose estimation algorithms for underwater manipulation | |
JP2001204757A (ja) | 人工眼球 | |
TW540010B (en) | Vision device | |
Nateghi | Detection, recognition and tracking cars from uav based implementation of mobilenet-single shot detection deep neural network on the embedded system by using remote sensing techniques | |
Shi et al. | A Review of Lane Detection Based on Semantic Segmentation | |
CN113487487B (zh) | 一种异构立体图像的超分辨率重建方法及系统 | |
Wan et al. | A cost-effective image sensor system for transport applications utilising a miniature CMOS single chip camera |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
WWE | Wipo information: entry into national phase |
Ref document number: 99810797.2 Country of ref document: CN |
|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): AE AL AM AT AU AZ BA BB BG BR BY CA CH CN CR CU CZ DE DK DM EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MD MG MK MN MW MX NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT UA UG US UZ VN YU ZA ZW |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): GH GM KE LS MW SD SL SZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF CG CI CM GA GN GW ML MR NE SN TD TG |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
DFPE | Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101) | ||
WWE | Wipo information: entry into national phase |
Ref document number: 56500/99 Country of ref document: AU |
|
WWE | Wipo information: entry into national phase |
Ref document number: 141589 Country of ref document: IL |
|
WWE | Wipo information: entry into national phase |
Ref document number: 1999943301 Country of ref document: EP |
|
ENP | Entry into the national phase |
Ref document number: 2000 570722 Country of ref document: JP Kind code of ref document: A |
|
WWE | Wipo information: entry into national phase |
Ref document number: 09786820 Country of ref document: US Ref document number: 1020017003096 Country of ref document: KR |
|
WWE | Wipo information: entry into national phase |
Ref document number: 1020017003103 Country of ref document: KR |
|
WWE | Wipo information: entry into national phase |
Ref document number: IN/PCT/2001/00288/MU Country of ref document: IN |
|
WWW | Wipo information: withdrawn in national office |
Ref document number: 1020017003096 Country of ref document: KR |
|
WWP | Wipo information: published in national office |
Ref document number: 1999943301 Country of ref document: EP |
|
REG | Reference to national code |
Ref country code: DE Ref legal event code: 8642 |
|
WWP | Wipo information: published in national office |
Ref document number: 1020017003103 Country of ref document: KR |
|
WWG | Wipo information: grant in national office |
Ref document number: 1020017003103 Country of ref document: KR |
|
WWG | Wipo information: grant in national office |
Ref document number: 56500/99 Country of ref document: AU |
|
WWG | Wipo information: grant in national office |
Ref document number: 1999943301 Country of ref document: EP |