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) 1. A method for enhancing the accuracy processing measurements of a physical quantity comprising the steps of:
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. 2. The method of claim 1 3. The method of claim 1 4. The method of 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. 5. The method of claim 1 6. The method of claim 2 7. A method for enhancing the accuracy of depth measurements of a portion of an ocean floor comprising the steps of:
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. 8. The method of claim 7 9. The method of claim 8 10. The method of claim 7 11. The method of 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. 12. The method of claim 7 13. A system for outputting anomalous measurements of an object of interest, comprising:
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. 14. The system of claim 13 15. The system of claim 13 16. The system of 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]FIG. 1 is an isometric drawing illustrating a representation of the floor of an ocean; [0011]FIG. 2 is a schematic drawing illustrating the measurement of the depth of a portion of the ocean floor; [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]FIG. 4 is a graph illustrating a normal distribution of residuals centered about a mean of zero; [0014]FIG. 5A is a flowchart illustrating a method for determining whether a measured value is anomalous; [0015]FIG. 5B is a graphical illustration of the method illustrated in FIG. 5A for determining whether a measured value is anomalous; [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]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]FIG. 1 is an isometric drawing illustrating a representation [0021]FIG. 2 is a schematic drawing illustrating the measurement of a depth [0022]FIG. 3 is a graph illustrating a plurality of measured values [0023] An outlier measurement [0024] In FIG. 4 there is shown a graph illustrating a normal distribution of residuals [0025] If outlier residual [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 [0027] The method begins at step [0028] A clear outlier is illustrated in FIG. 5B as producing clear outlier data set [0029] One example of a procedure for temporarily removing clear outliers is described below. A standard moving window technique in which measured values [0030] After clear outliers have been temporarily removed, the trend of data set [0031] After data set [0032] 10*Σ [0033] [0034] where Σ [0035] Step [0036] As described above, the data set [0037] After generating residuals [0038] Non-outlier residuals [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 [0040] Based on the computed variogram [0041] At a step | [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 [0048]FIG. 5C is a plot of the variogram [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 [0050] In the described example of selecting whether an outlier residual is anomalous or non-anomalous, the measured values of depth [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]FIG. 6 is a block diagram of anomalous measurement detection system [0053] Anomalous measurement detection system [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
Classifications
Legal Events
Rotate |