Publication number | US20010004726 A1 |

Publication type | Application |

Application number | US 09/776,189 |

Publication date | Jun 21, 2001 |

Filing date | Feb 1, 2001 |

Priority date | Oct 13, 1998 |

Also published as | US6185512, US6351717 |

Publication number | 09776189, 776189, US 2001/0004726 A1, US 2001/004726 A1, US 20010004726 A1, US 20010004726A1, US 2001004726 A1, US 2001004726A1, US-A1-20010004726, US-A1-2001004726, US2001/0004726A1, US2001/004726A1, US20010004726 A1, US20010004726A1, US2001004726 A1, US2001004726A1 |

Inventors | Mark Lambrecht |

Original Assignee | Raytheon Company |

Export Citation | BiBTeX, EndNote, RefMan |

Referenced by (13), Classifications (9), Legal Events (5) | |

External Links: USPTO, USPTO Assignment, Espacenet | |

US 20010004726 A1

Abstract

A method for determining measurements of a surface includes determining a trend for a plurality of measurements and determining, for each of at least one of the plurality of measurements, an associated residual between the trend and the measurement and determining, for each of at least one of the associated residuals, an interpolated value for the residual. The method also includes comparing, for each of the at least one associated residuals, the residual to the interpolated value of the residual to determine whether a measurement associated with the residual is anomalous. In response to determining that a measurement associated with the residual is anomalous, the anomalous measurement is discarded. The remaining measurements then represent the surface of interest. The system includes a processor for executing a program of the method. Further, the system includes memory, an input device, and an output device.

Claims(16)

determining a trend for a plurality of measurements;

determining, for each of at least one of the plurality of measurements, an associated residual between the trend and the measurement;

determining, for each of at least one of the associated residuals, an interpolated value for the residual;

comparing, for each of the at least one associated residuals, the associated residual to the interpolated value of the residual to identify an anomalous associated measurement;

in response to identification of an anomalous measurement, discarding the anomalous measurement; and

generating an output of all measurements of the physical quantity not discarded as anomalous.

claim 1

claim 1

claim 1

determining, for each of the at least one of the plurality of measurements, a residual error associated with the residual;

determining, for each of the at lest one of the associated residuals, an interpolated value error associated with the interpolated value; and

wherein the step of comparing, for each of the at least one associated residuals, the associated residual to the interpolated value of the residual comprises determining whether the magnitude of the difference between the associated residual and the interpolated value of the residual exceeds {square root}{square root over (δI^{2}+δD^{2})}, where “δI” is the interpolated value error and “δD” is the residual error.

claim 1

claim 2

determining a trend for a plurality of depth measurements;

determining, for each of the plurality of depth measurements, an associated residual between the trend and the measurement;

determining when the magnitude of the residual exceeds a particular value;

determining, for each of the residuals that exceeds the particular value, an interpolated value for the residual;

comparing, for each of the residuals that exceeds the particular value, the residual to the interpolated value of the residual to identify an anomalous associated depth measurement;

in response to identification of an anomalous associated depth measurement, discarding the anomalous depth measurement; and

generating an output of all depth measurement of the physical quantity not discarded as anomalous.

claim 7

claim 8

claim 7

claim 10

determining for each of the residuals that exceeds a particular value, an error value associated with the determined interpolated value for the residual;

determining, for each of the plurality of depth measurements, a residual error associated with the residual; and

wherein the step of comparing, for each of the residuals that exceeds the particular value, the residual to the interpolated value of the residual includes the step of determining whether the difference between the residual and the interpolated value of the residual exceeds {square root}{square root over (δI^{2}+δD^{2})}, where “δI” is the interpolated value error and “δD” is the residual error.

claim 7

a storage medium;

a processor coupled to the storage medium;

a computer program stored in the storage medium, the computer program executed on the processor to:

determine a trend for a plurality of measurements;

determine, for each of at least one of the plurality of measurements, an associated residual between the trend and the measurement;

determine, for each of at least one of the associated residuals, an interpolated value for the residual;

compare, for each of the at least one associated residuals, the error value to the residual to determine whether a measurement associated with the residual is anomalous;

in response to an identification of an anomalous measurement, discarding anomalous measurements; and

output means for generating all measurements of the object of interest not discarded as anomalous.

claim 13

claim 13

claim 13

determining, for each of the at least one of the plurality of measurements, a residual error associated with a residual;

determining for each of the at least one of the associated residuals an interpolated value error associated with the interpolated value; and

determines, for each of the at least one associated residuals, whether the magnitude of the difference between the residual and the interpolated value of the residual exceeds {square root}{square root over (δI^{2}+δD^{2})}, where “δI” is the interpolated value error and “δD” is the residual error.

Description

- [0001]This invention relates generally to enhancing the accuracy of measurements of a physical quantity and more particularly to a method and system for enhancing the accuracy of measurements of a physical quantity.
- [0002]Modeling of a physical object often requires the taking of measurements. The measurement taking process on occasion introduces an error associated with each measured value. If the measurement taking process is conducted in an appropriate fashion, the errors should be normally distributed about a mean of zero. However, at times, the measurement taking or measurement recordation process is flawed, resulting in flawed or anomalous data. In such a case the use of such data to represent the measured object would be incorrect and would provide imprecise results. Therefore, it is desirable to determine whether a measurement is anomalous before using the measurement to represent a physical quantity.
- [0003]Although anomaly detection has utility in numerous applications and for numerous types of measurements, one example of the usefulness of anomaly detection will be described with reference to bathymetric data. Oceanographers and ocean explorers make use of charts of the ocean floors. Such charts are produced by taking depth measurements of the ocean floor. One example of a procedure for taking a depth measurement of the ocean floor involves projecting, from an ocean vessel, sound waves to the bottom of the ocean floor and measuring the time required for a reflected wave to return to the vessel. Such a procedure is subject to a variety of sources of error that create anomalous measurements. For example, a large ocean animal, such as a whale, may reflect a portion of the sound wave back to the vessel, creating an anomalous reading. In addition, measurements may be improperly recorded in meters where feet is the appropriate unit of measure, or feet where meters are the appropriate unit of measure. Such sources of errors lead to anomalous measurements that should be discarded prior to utilizing the data to represent a physical quantity.
- [0004]Accordingly, a need has arisen for an improved method and system for enhancing the accuracy of measurements of a physical quantity. The present invention provides a system and method for enhancing the accuracy of measurements representing a physical quantity.
- [0005]According to one embodiment of the invention, a method for enhancing the accuracy of measurements representing a physical quantity includes determining a trend for a plurality of measurements. For each of at least one of the plurality of measurements an associated residual between the trend and the measurement is determined, and, for at least one of the associated residuals, there is determined an interpolated value for the residual. The method also includes comparing, for each of the at least one associated residuals, the residual to the interpolated value of the residual to determine whether a measurement associated with the residual is anomalous. In response to determining that a measurement associated with the residual is anomalous, the anomalous measurement is discarded. The remaining non-anomalous measurements then represent the physical quantity.
- [0006]According to another embodiment of the invention, a system for enhancing the accuracy of measurements of a physical quantity includes a storage medium, a processor coupled to the storage medium, and a computer program stored in the storage medium. The processor executes the computer program to determine a trend for a plurality of measurements. The processor executing the computer program generates a determination, for each of at least one of the plurality of measurements, an associated residual between the trend and the measurement, and also determines, for each of at least one of the associated residuals, an interpolated value for the residual. The processor running the computer program also compares, for each of the at least one associated residuals, the interpolated value to the residual to determine whether a measurement associated with the residual is anomalous. In response to determining whether a measurement is anomalous, execution of the computer program generates an indication that the measurement is anomalous so that it may be discarded. The non-anomalous measurements are then output from the system.
- [0007]Embodiments of the invention provide numerous technical advantages. For example, in one embodiment of the invention, a method for enhancing the accuracy of measurements allows determination of whether a measurement is anomalous and whether it may be discarded. Discarding anomalous measurements creates a more accurate representation of the measured object than if the anomalous measurements were included. In one embodiment of the invention a Kriging process is utilized to perform an interpolation used in determining if a measurement is anomalous. The use of the Kriging process is particularly advantageous because Kriging automatically provides an error for an interpolated value, which is utilized in the present invention and which is not readily available with other interpolation methods without additional processing.
- [0008]Other technical advantages are readily apparent to one skilled in the art from the following figures, descriptions, and claims.
- [0009]For a more complete understanding of the present invention and the advantages thereof, reference is now made to the following descriptions taken in connection with the accompanying drawings in which:
- [0010][0010]FIG. 1 is an isometric drawing illustrating a representation of the floor of an ocean;
- [0011][0011]FIG. 2 is a schematic drawing illustrating the measurement of the depth of a portion of the ocean floor;
- [0012][0012]FIG. 3 is a graph illustrating the plurality of measured values of the depth of the ocean floor at different longitudes as well as the actual depth of the ocean floor at those longitudes;
- [0013][0013]FIG. 4 is a graph illustrating a normal distribution of residuals centered about a mean of zero;
- [0014][0014]FIG. 5A is a flowchart illustrating a method for determining whether a measured value is anomalous;
- [0015][0015]FIG. 5B is a graphical illustration of the method illustrated in FIG. 5A for determining whether a measured value is anomalous;
- [0016][0016]FIG. 5C is a plot of a variogram used according to the method of FIG. 5A in determining whether a measurement is anomalous; and
- [0017][0017]FIG. 6 is a block diagram of an anomalous measurement detection system.
- [0018]The present invention and advantages thereof are best understood by referring to FIGS. 1 through 6 of the drawings, like numerals being used for like and corresponding parts of the various drawings.
- [0019]Although the invention is useful to determine anomalies in measurements in a variety of contexts, an example application for the invention is the determination of whether depth measurements of the ocean floor are anomalous. Depth measurements of the ocean floors are often referred to as bathymetric data. The following description of the invention is in the context of determining anomalies in depth measurements of the oceans; however, the invention also finds utility with any suitable type of measurements including, but not limited to, measurement of temperature, pressure, volume, weight, mass, length, area, velocity, acceleration, luminosity, energy, power, electrical charge, and magnetism.
- [0020][0020]FIG. 1 is an isometric drawing illustrating a representation
**10**of an ocean floor**16**. Representation**10**provides a model of ocean floor**16**, which is similar to models used by oceanographers or other persons interested in the shape and depth of ocean floor**16**. The depth of ocean floor**16**is illustrated at a variety of locations**12**of ocean floor**16**. Locations**12**are specified by a longitude**26**and a latitude**28**. Such a model is developed through measurements of the ocean floor taken according to one method as described in conjunction with FIG. 2. According to the invention, such measurements are analyzed in order to exclude anomalous measurements for use in generating representation**10**. Exclusion of anomalous measurements generates a more accurate representation of a physical quantity. - [0021][0021]FIG. 2 is a schematic drawing illustrating the measurement of a depth
**14**of ocean floor**16**at a location**12**. As illustrated, a vessel**20**takes depth measurements by transmitting sound waves**22**to the ocean floor**16**. By measuring the time for sound waves**22**to reflect back to vessel**20**, an indication of depth**14**of ocean floor**16**is obtained. A location**12**at which depth**14**is measured is specified by longitude**26**and latitude**28**. These depth measurements are subject to errors in the measurement taking process and should be examined for anomalous measurements to produce the best possible representation**10**of the ocean floor**16**. For example, a large ocean animal**24**, such as a whale, may alter the reflection time of sound waves**22**to return to vessel**20**and therefore produce a measurement that is not accurately indicative of depth**14**. Such a measurement would be anomalous. Another example of the production of an anomalous measurement is the recordation of depth**14**in feet if the appropriate unit of measure is meters. - [0022][0022]FIG. 3 is a graph illustrating a plurality of measured values
**30**of depth**14**of ocean floor**16**at different longitudes**26**as well as the actual depth**32**of the ocean floor at those longitudes**26**. A dotted line**56**represents a calculated trend of measured depths**30**. Calculation of trend line**56**of measured depth**30**is described in greater detail in conjunction with FIGS. 5A through 5C. Although in the example illustrated in FIG. 2 measured depths**30**are taken at a given longitude**26**and latitude**28**, for simplicity of description FIG. 3 illustrates varying depths as a function of longitude**26**for a constant latitude**28**. As illustrated, most measured depths**30**lie close to trend line and are located on either side of the trend line. Also illustrated in FIG. 3 is one of a plurality of residuals**38**. Each residual**38**indicates a difference between a measured depth**30**and trend line**56**. - [0023]An outlier measurement
**34**is illustrated as lying significantly below the trend line and has an associated outlier residual**39**. An outlier measurement is a measurement that differs significantly from a trend of the measured values, such as trend line**56**. An outlier measurement may be anomalous or non-anomalous. An anomalous outlier measurement is an outlier measurement that is determined not to be representative of the actual physical quantity. A non-anomalous outlier measurement is an outlier measurement that, although differing from a determined trend of measured values, is nevertheless determined to be representative of the actual quantity measured. Determination that a measurement is an outlier measurement**34**is performed according to the teachings of the invention as described with reference to FIGS. 5A through 5C. - [0024]In FIG. 4 there is shown a graph illustrating a normal distribution of residuals
**38**centered about a mean of zero. As illustrated, a normal bell-shaped distribution curve**36**represents an expected distribution of errors for measurements taken, for example, according to the method illustrated in FIG. 2. A normal bell-shaped distribution curve**36**is centered about a mean of zero, thereby indicating that the number of data values**30**exceeding trend line is approximately equal to the number of measured values**30**less than trend**56**. Outlier residual**39**is illustrated on the far left hand side of the distribution curve**36**. Such a residual value may lie on the far end of the curve because it is anomalous, and indicate that the measurement was improperly taken. Alternatively, the residual**39**may be at the far end of the curve because it is consistent with the normal distribution curve**36**, but nevertheless is an extreme residual value. - [0025]If outlier residual
**39**occurred because of an anomalous measurement, it would be desirable to discard outlier measurement**34**associated with outlier residual**39**to avoid using an anomalous measurement in producing representation**10**of ocean floor**16**. If, however, outlier residual**39**was obtained correctly and merely corresponds to an extreme value on normal distribution curve**36**, outlier measurement**34**should be retained in producing representation**10**of ocean floor**16**. Determination of whether outlier measurement**34**is anomalous is made in accordance with the present invention as described with reference to FIGS. 5A through 5C. - [0026][0026]FIG. 5A is a flowchart illustrating a method for determining whether a measured value is anomalous, and FIG. 5B is a graphical illustration of the method of FIG. 5A for determining whether a measured value is anomalous. A method for determining whether an outlier is an anomalous measurement, such as outlier measurement
**34**, is described in conjunction with FIGS. 5A and 5B. For simplicity of description, the method is described in the context of bathymetric data obtained as illustrated in FIG. 2; however, as described above, the method may be used for a variety of types of measurements for various physical quantities. - [0027]The method begins at step
**40**illustrated in FIG. 5A. At a step**42**a source data set**142**illustrated in FIG. 5B, is received. Source data set**142**includes a plurality of measured values**30**of depth**14**at associated longitudes**26**and latitudes**28**. At a step**44**clear outliers, such as outlier measurement**34**, are temporarily removed from the processed data to produce a clear outlier as data set**146**and a data set**148**that does not include clear outliers as illustrated in FIG. 5B. - [0028]A clear outlier is illustrated in FIG. 5B as producing clear outlier data set
**146**and the data set**148**to facilitate description of the present invention; however, such illustration does not imply that a separate device or program is necessarily utilized for performing step**44**. Additional representations in FIG. 5B, such as, a residual calculator**158**, an outlier detector**164**, a variogram calculator**174**, and an outlier classifier**180**are similarly illustrated to facilitate description, and their functions are similarly not necessarily performed by independent devices or programs. - [0029]One example of a procedure for temporarily removing clear outliers is described below. A standard moving window technique in which measured values
**30**are binned at, for example, 20 times the measured Real World Nyquist of data measurements is used. Binning refers to the generation of a regular array of map cells overlying the source data set and is used to compute statistics of data points lying in a small geographical area. A Real World Nyquist is the frequency beyond which no further information may be gleaned from the data. A mean and a standard variation for measured values**30**is calculated for each bin. Measured values**30**that exceed a predetermined multiple of the standard deviation of measured values**30**are temporarily removed as clear outliers. It has been determined that temporarily excluding errors that exceed five times the standard deviation is particularly advantageous; however, other suitable multiples of the standard deviation may be utilized to remove clear outliers. Step**44**of temporarily removing clear outliers is performed to avoid undue influence on a subsequent step**50**for computing a trend of measured values**30**. - [0030]After clear outliers have been temporarily removed, the trend of data set
**148**is determined at step**50**. A preliminary step in computing the trend of data set**148**is to grid data set**148**at step**52**. Gridding data set**148**produces a grid of data points that have values representing the depths at each grid point. These values at each grid point are based upon data within data set**148**. An example technique for gridding data set**148**is the use of the Triangular Irregular Network algorithm. The Triangular Irregular Network algorithm is described in Macedonio, G. and Pareschi, M. T., 1991,*An Algorithm for the Triangulation of Arbitrarily Distributed Points: Applications to Volume Estimate and Terrain Fitting,*Computers and Geosciences, v. 17, no. 17, p. 859-874. - [0031]After data set
**148**has been gridded, the gridded data is filtered at step**54**to maintain only the low frequency components of the gridded data. The low frequency components of the gridded data represent the trend of the gridded data. Step**54**of filtering the gridded data is illustrated in FIG. 5B as being performed by a low pass filter function**154**. In one embodiment of the invention, the filtering at step**54**utilizes a simple Gaussian expression with a half-width given by: - [0032]10*Σ
_{d}; - [0033][0033]
- [0034]where Σ
_{d }is the horizontal resolution of the input data computed by a spectral analysis. This corresponds to maintaining wavelengths of the data greater than 10 times the Real World Nyquist of the data. - [0035]Step
**54**of filtering the gridded data produces a low pass component or trend**156**illustrated in FIG. 5B. Low pass component or trend**156**represents the trend of data set**148**. An example illustration of a low pass component is provided in one dimension for a constant latitude as trend**56**in FIG. 3. Based on the computed low pass component or trend**156**of data set**148**, the difference between trend**156**and each data point of source data set**142**is calculated at step**58**. This difference between each data point and the calculated trend is referred to as a residual. Step**58**of computing residuals is illustrated in FIG. 5B as being performed by a residual calculator**158**. As shown, the residual calculator**158**receives data values from both clear outlier data set**146**and data set**148**. Thus clear outliers in data set**146**are temporarily removed to predict trend**156**of source data**142**, but are later used after trend**156**is determined. - [0036]As described above, the data set
**148**is gridded before the low pass component or trend**156**is calculated, and as a result the low pass component or trend**156**must also be gridded. Therefore, some interpolation is necessary to determine a measurement at a random location on the gridded surface of low pass component or trend**156**to provide an associated residual. This may be accomplished according to a variety of techniques including simple bilinear interpolation. Residuals are calculated by subtracting the data values both in clear outlier data set**146**and data set**148**, or in other words the data values from source data set**142**, from the gridded low pass component or trend**156**. A set of residuals**160**illustrated in FIG. 5B is produced by step**58**that computes residuals for all data within source data set**142**. - [0037]After generating residuals
**160**at step**58**, the mean and standard deviation of each residual of the set of residuals**160**is calculated at a step**62**. At a step**64**each of the residuals are compared to the standard deviation calculated at step**62**. If a residual exceeds a particular multiple of the standard deviation of residuals, the residual is identified as an outlier, and the residual may be anomalous or non-anomalous. If a residual is less than a particular multiple of the standard deviation of the residuals, the residual is identified as a non-outlier. In one embodiment of the invention, a multiple of 2.5 has been determined to provide particularly advantageous results. Steps**62**and**64**of computing the mean and standard deviation for each of the residuals**160**, comparing residuals to the standard deviation of residuals**160**, and selecting outliers is illustrated in FIG. 5B as performed by an outlier detector**164**. The results of these steps produce a set of outlier residuals**166**and a set of non-outlier residuals**168**. - [0038]Non-outlier residuals
**168**do not differ from the mean of all of the residuals**160**to an extent great enough to constitute anomalous data. Outlier, residuals**166**differ from the mean of residuals**160**to an extent great enough to be anomalous. Step**64**of comparing residuals**160**to the standard deviation of residuals**160**and selecting outliers is performed to increase processing speed of anomalous data detection. Alternatively, step**64**is omitted and the determination of whether a residual is associated with an anomalous measurement is performed on each residual of the set of residuals**160**rather than only residuals**166**that constitute outliers. The remainder of the method illustrated in FIGS. 5A and 5B is directed to determining whether outlier residuals**166**are anomalous or non-anomalous. - [0039]One method for differentiating between residuals that are anomalous and residuals that are not anomalous is cross validation. Cross validation differentiates between residuals that are supported by neighboring depth measurements and those that are not. Residuals supported by their neighbors, despite being abnormally high, are much more likely to be valid data points than isolated outliers. Cross validation begins at step
**70**. A first step in cross validation occurs at a step**72**in which residuals**160**are gridded. Gridding residuals**160**is performed in the same manner as described above with reference to gridding data set**148**. At a step**74**, variograms are calculated for gridded residuals**160**. Variograms are used to determine which neighboring measurements should be used in determining whether a residual is supported by its neighbors and facilitates interpolation of a data value based on neighboring data values. Variograms are described in greater detail in conjunction with FIG. 5C. Steps**70**and**72**of gridding residuals**160**and computing a variogram for residuals**160**are illustrated in FIG. 5B for simplicity of description as being performed by a variogram calculator**174**. Variograms**176**, illustrated in FIG. 5B, are produced by step**74**. - [0040]Based on the computed variogram
**176**for each outlier residual, an interpolated value is calculated at step**76**for each of the outlier residuals**166**based upon the value of neighboring residuals to determine whether an outlier residual is anomalous or non-anomalous. For this calculation, the value calculated for an outlier residual at step**58**is not included. Thus an outlier residual in the set of outlier residuals**166**is not included in the determination of an interpolated value for that outlier residual. In addition to determining an interpolated value for an outlier residual, an error associated with the interpolated value is determined. One example of step**76**of determining an interpolated residual value and error is described in greater detail in conjunction with FIG. 5C. - [0041]At a step
**78**, the interpolated value for an outlier residual is compared to the measured value of the same outlier residual to determine if they are statistically inconsistent. If the measured value of an outlier residual differs from the interpolated value of an outlier to too great a degree an outlier residual is classified as anomalous. If however, an outlier residual in the set of outlier residuals**166**differs from the interpolated value of an outlier residual by less than an acceptable amount, the outlier residual is classified as non-anomalous. In one embodiment, the determination of whether the interpolated value for an outlier residual differs from the measured value of an outlier residual is made according to the formula - |
*D−I|>{square root}{square root over (δI*^{2}+δD^{2})} - [0042]where
- [0043]I=interpolated value for a residual
- [0044]D=measured value for a residual
- [0045]δI=calculated error for I
- [0046]δD=calculated error for D.
- [0047]The calculated error for the measured value of residuals (δD) is determined based on the error in the measurement of the associated depth. If the left-hand-side of the above equation exceeds the right-hand-side, the associated residual is classified as anomalous. This classification is performed at step
**80**. Steps**76**,**78**, and**80**are illustrated in FIG. 5B, to facilitate description, as being performed by an outlier classifier**180**. The results of step**80**are a data set**182**of non-anomalous outliers and a data set**184**of anomalous outliers. Data sets**182**and**184**are generated at a step**86**. At a step**88**data in source data set**142**associated with residuals in anomalous outlier data set**184**are discarded. In addition, data in source data set**142**associated with residuals in non-anomalous data set**182**are combined with data in source data set**142**associated with residuals in non-outlier data set**168**for use in modeling the measured object. The method concludes at step**90**. - [0048][0048]FIG. 5C is a plot of the variogram
**176**used according to the method of FIG. 5A in determining whether a measurement is anomalous. Variogram**176**identifies the extent an outlier residual**178**is correlated with data values proximate the outlier residual**178**. Neighboring data values are correlated with outlier residual**178**if they lie within the range of the variogram centered at the residual under consideration. Therefore, only data values within variogram**176**are utilized to calculate an interpolated value for outlier residual**178**. The interpolation process determines an interpolated value for outlier residual**178**based on the residual values of neighbor residuals**202**within variogram**176**using a weighted average. The weighting is based, in part, on the extent to which data values within variogram**176**are correlated with outlier residual**178**. This correlation varies roughly with the distance between a neighboring residual**202**and outlier residual**178**. In addition, the weighting is dependent upon the uncertainty of the residual values of neighboring residuals**202**within variogram**176**. - [0049]An exemplary process of implementing such interpolation and calculation of a variogram is referred to in the art as multi-variate Kriging and is described in
*Introduction to Disjunctive Kriging and Nonlinear Geostatistics,*by J. Rivoirand, 1994, Clarendon Press, Oxford University Press, which is incorporated by reference herein. Although an interpolated value for outlier residual**178**and an associated error may be obtained according to other interpolation methods, the use of Kriging is particularly advantageous. - [0050]In the described example of selecting whether an outlier residual is anomalous or non-anomalous, the measured values of depth
**14**were obtained on ocean surface**16**for various longitudes**26**and latitudes**28**. Thus, interpolation utilized neighbors lying roughly in a surface specified by two dimensions. However, other applications for the invention utilize interpolation in a single dimension and in more than two dimensions. - [0051]Thus, the method described in conjunction with FIGS. 5A through 5C determines whether a measurement is anomalous or non-anomalous and allows discarding of anomalous measurements in order to provide a more accurate representation of the measured object. In one embodiment of the invention, Kriging is used to perform an interpolation and error calculation. Kriging is particularly useful for these functions because Kriging automatically provides an error with the interpolated value, alleviating the need to perform additional calculations that would be required if an interpolation method was used that did not automatically provide an error calculation. An example of that utilizes the above method is described below in conjunction with FIG. 6.
- [0052][0052]FIG. 6 is a block diagram of anomalous measurement detection system
**100**for determining whether a measured value is anomalous. Anomalous measurement detection system**100**includes a processor**202**, an input device**202**, an output device**206**, a memory system**108**, and a disk drive**110**. The present embodiment includes computer software stored in memory system**108**or on disk drive**110**and is executed by processor**202**. Disk drive**110**includes a variety of types of storage media such as, for example, floppy disk drives, hard drives, CD ROM disk drives, or magnetic tape drives. Data is received from a user of anomalous measurement detection system**100**using a keyboard or other type of input device**204**. Data is provided to a user of anomalous data detection system**100**through output device**206**, which may include a display, printer, or any other suitable type of output device. - [0053]Anomalous measurement detection system
**100**includes an anomaly detection program**112**. In FIG. 6, the anomaly detection program**112**is illustrated as stored in memory system**108**, and will be executed by processor**202**. Anomaly detection program**112**is alternatively stored on disk drive**110**. The processor**202**when executing the anomaly detection program**112**receives source data set**142**and determines which data values within the source data set**142**are anomalous and which are non-anomalous according to the method described in conjunction with FIGS. 5A through 5C. The system provides an outlier representing the data values that are anomalous and the data values that are non-anomalous. The anomalous measurement data detection system**100**discards anomalous data to enhance the accuracy of measurements representing a physical quantity. - [0054]Although the present invention and its advantages have been described in detail, it should be understood the various changes, substitutions, and alterations can be made therein without departing from the spirit and scope of the present invention as defined by the appended claims.

Referenced by

Citing Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|

US6816811 * | Jul 20, 2001 | Nov 9, 2004 | Johnson Controls Technology Company | Method of intelligent data analysis to detect abnormal use of utilities in buildings |

US7523396 | Sep 29, 2004 | Apr 21, 2009 | Autodesk, Inc. | Surface construction audit trail and manipulation |

US7561990 | Jul 14, 2009 | Autodesk, Inc. | Interactive triangulated irregular network (TIN) surfaces design | |

US8065124 * | Jul 20, 2008 | Nov 22, 2011 | China Institute Of Water Resources And Hydropower Research | Method for predicting micro-topographic distribution of terrain |

US9146903 | Nov 15, 2011 | Sep 29, 2015 | Chevron U.S.A. Inc. | Method of using spatially independent subsets of data to calculate vertical trend curve uncertainty of spatially correlated reservoir data |

US9171339 * | Nov 3, 2011 | Oct 27, 2015 | International Business Machines Corporation | Behavior change detection |

US20030028350 * | Jul 20, 2001 | Feb 6, 2003 | Seem John E. | Method of intelligent data analysis to detect abnormal use of utilities in buildings |

US20050068315 * | Sep 29, 2004 | Mar 31, 2005 | Autodesk, Inc. | Surface construction audit trail and manipulation |

US20050078110 * | Sep 29, 2004 | Apr 14, 2005 | Autodesk, Inc. | Interactive triangulated irregular network (TIN) surfaces design |

US20050091016 * | Sep 29, 2004 | Apr 28, 2005 | Autodesk, Inc. | Surface smoothing techniques |

US20090024371 * | Jul 20, 2008 | Jan 22, 2009 | Xu Di | Method for predicting micro-topographic distribution of terrain |

US20130116939 * | May 9, 2013 | International Business Machines Corporation | Behavior change detection | |

WO2013074157A1 * | Jul 25, 2012 | May 23, 2013 | Chevron U.S.A. Inc. | Method of using spatially independent subsets of data to calculate vertical trend curve uncertainty of spatially correlated reservoir data |

Classifications

U.S. Classification | 702/5 |

International Classification | G01S15/89, G01S7/52 |

Cooperative Classification | G01S7/52004, G01S15/89, G01S7/53 |

European Classification | G01S7/53, G01S7/52G, G01S15/89 |

Legal Events

Date | Code | Event | Description |
---|---|---|---|

Jul 13, 2005 | FPAY | Fee payment | Year of fee payment: 4 |

Aug 20, 2009 | FPAY | Fee payment | Year of fee payment: 8 |

Oct 4, 2013 | REMI | Maintenance fee reminder mailed | |

Feb 26, 2014 | LAPS | Lapse for failure to pay maintenance fees | |

Apr 15, 2014 | FP | Expired due to failure to pay maintenance fee | Effective date: 20140226 |

Rotate