JP6509887B2 - Image change detection - Google Patents

Description

  The present application relates to color image processing, and more particularly to image segmentation to determine a list of ranked segments according to the likelihood of change.

  Property taxes are an important revenue raising tool for local governments, such as cities and towns. Tax assessment is a way to assign property taxes between individuals who own real estate in a municipality. Tax allocation using the above valuation method generally works according to the following process. First, each parcel of real estate within the municipality is "rated"-that is, assigned a relative value for the limited purpose of collecting property taxes. Next, the total valuation value for the municipality is calculated by adding up the valuation values for all the parcels. Finally, the property tax to be collected for each section is allocated by multiplying the ratio of the total property tax to be raised to the total assessed value of each respective section. As can be understood from the present specification, the valuation property value does not have to be related to the market value of the property as long as all the valuation values are appropriately determined to each other. Thus, the valuation value represents the relative value of the parcel of real estate relative to its neighbors.

  Allocation fairness in tax burden requires that parcels be reassessed periodically to capture increases and decreases in their relative value. However, it is difficult and expensive to determine which parcels have been improved and the degree of improvement, especially if done by an individual tax evaluator who physically sees all the parcels in the municipality. And, it may be a time consuming task. One tool that can simplify the process is remote sensing. Remote sensing, in the present context, means a means of acquiring an aerial image or satellite image of a municipality multiple times. The tax assessor compares the "previous" image captured at the time of the previous tax assessment with the more recently captured "after" image against which the assessment has to be performed. It may be determined which property has improved over the relevant evaluation period. Side-by-side visual image comparison can be a useful tool, but many compartments (especially in the case of developmental or semi-developmental compartments laid out to provide a uniform appearance) ) Because they look almost the same, they can be painfully mentally burdened. Thus, automatic image analysis may be employed.

  Image analysis is a process for extracting meaningful information from an image. One type of image analysis that is particularly useful in the present context is geographical object-based image analysis (GEOBIA), among others. Digital images are formed from pixels, and GEOBIA takes those pixels and performs two main processes: segmentation and classification. Image segmentation is the process of partitioning a digital image into segments, each of which is formed of a plurality of pixels having common characteristics, such as color or luminance. Such segments thus represent the type of land cover (i.e. grass, asphalt, trees, soil, water, etc.). Statistics may then be applied to the segments to classify each segment according to the type of land cover it is representing.

  However, image analysis may extract meaningful information, including changes to the compartments that affect assessment value in the context of remote sensing (eg, addition of a solar cell system on the roof, or addition of a swimming pool on the ground) Required Meaningful information does not include changes between images caused by transients (eg, shadows, typically differences in light and dark, the presence or absence of humans, cars, other moving objects, etc.). Distinguishing these types of changes is generally a very difficult problem, and existing segmentation and classification software does not make sense segments meaningful, especially in the context of performing tax assessments. It is plagued by various problems of distinction from

(Summary of the embodiment)
In accordance with various embodiments of the present invention, systems, methods, and computer program products compare images of physical activities or objects. Pixel data from the first registered color image and the second registered color image defined in the N-dimensional color model are concatenated to form an image in the 2N-dimensional color model. Principal component analysis is performed on the composite image to identify components that represent changes rather than correlations. In the composite image, the coordinates for each pixel are determined along the main component that represents the change, and normalized pixel change values are calculated. In particular, in connection with performing a tax assessment, the pixels in the composite image are segmented according to their respective normalized pixel change values, and the segments do not mutually predefine geographic parcel boundaries . The parcel change probabilities are calculated as a function of segment change probabilities (ie, their function on normalized pixel change values). Finally, the compartments are ranked by their probability of change and highlighted in the graphical user interface.

  Thus, a first embodiment of the present invention provides a method for comparing a first digital color image and a second digital color image having aligned pixels, each pixel having data for N colors. And the first digital color image has the physical activity or the appearance of the object at a first point in time, and the second digital color image the physical activity or the appearance of the object at a second point in time Have. The method has four processes. The first process defines the color data for each pixel in the composite image as a concatenation of the color data for pixels in the first digital color image and the color data for aligned pixels in the second digital color image. Is to form a composite image having pixels in a 2N dimensional color model. The second process is to determine the set of main components of the composite image using main component analysis, wherein the set of main components define a coordinate system in a 2N dimensional color model. The third process takes a subset of the main components in the coordinate system as main components each representing differences, rather than correlations between the first digital color image and the second digital color image. It is to choose. The fourth process is to calculate, for each image in the composite image, normalized pixel change values as a function of the pixel color data coordinates along each main component in the subset, The pixel change values collectively define a change image having a one-dimensional color model, and the calculated normalized pixel change values for each pixel are physically between the first and second time points. Changes in physical activity or objects.

  Various refinements of the basic method are envisaged. In one embodiment, N = 3, and each pixel has red, green and blue (RGB) color data. In another embodiment, N = 4, and each pixel has red, green, blue, and infrared (RGB IR) color data. In yet another embodiment, selecting the subset of main components comprises, for each main component of the composite image, (a) from a second N-dimensional portion of the main components representing a second digital color image Forming an N-dimensional difference vector by subtracting a first N-dimensional portion of the main component representing the first digital color image, (b) an N-dimensional length of the difference vector being of a given length And if so, selecting the main component for inclusion in the subset. In still another embodiment, the normalized pixel change value for each pixel is calculated as the maximum of the color data coordinates of the pixel along each main component in the subset. Yet another embodiment calls for removing pixels with intensities below the shadow threshold from the composite image before determining the set of main components of the composite image.

  The method may in particular be applied to perform a tax assessment. In this case, the physical activity or object is a municipality, and the first digital color image represents the appearance of the municipality at a first time point, and the second digital color image is of the municipality at a second time point. In appearance, a municipality is divided into a plurality of compartments. The method may further include four additional processes. The first additional process is to divide the change image into a plurality of segments (a) each given segment in the plurality of segments is formed from consecutive pixels and completely within the compartments in the plurality of compartments And (b) the membership of each pixel in a given segment is determined as a function of its normalized pixel change value. The second additional process is to calculate, for each segment in the plurality of segments, the segment change probability as a function of the normalized pixel change value of the pixels forming it. A third additional process is to calculate, for each partition in the plurality of partitions, the partition change probability as a function of the segment change probability of the segments within it. A fourth additional process is to rank the compartments in the plurality of compartments according to their respective compartment change probabilities.

  Various refinements of tax assessment applications are envisioned. In one embodiment, the first digital color image and the second digital color image represent the appearance of a municipality approximately integer years apart. In another embodiment, the segment change probability for each segment in the plurality of segments is calculated as the average of the normalized pixel change values of the pixels that form it. Yet another embodiment calls for excluding segments with segment change probabilities less than the segment change threshold before computing the segment change probability. A further embodiment calls for excluding segments consisting of a number of pixels representing an area having a number of square feet less than the segment size before computing the parcel change probability. In still another embodiment, the compartment change probability for each compartment in the plurality of compartments is the segment change probability for a given segment, the number of pixels in a given segment, and the number of pixels in a compartment containing a given segment Calculated as the sum of the product of the ratios between and over each segment within a given partition.

  The tax assessment method further includes providing a graphical user interface (GUI) to the computer system, the GUI displaying a portion of the first digital color image next to the corresponding portion of the second digital color image. The display on the allows a visual comparison of parts by the user of the computer system. In this case, the method also causes the GUI to (a) highlight the contours of the compartments in a portion of the first digital color image and (b) highlight the compartments in a corresponding portion of the second digital color image. It may include simultaneously displaying the contour and (c) a section with significant change, a section without significant change, or a control to receive a selection indicating a section requiring further analysis. Alternatively or additionally, the method may include providing a control that allows the user to adjust the shadow threshold, the segment change threshold, or the segment size threshold.

  Also envisioned is a non-volatile, tangible computer-readable storage medium having stored thereon computer program code for performing any of the above methods or refinements thereof.

Further, a computer system is envisioned having volatile memory having stored therein computer program code for performing any of the above methods or refinements thereof. The computer system also provides a microprocessor coupled to the volatile memory and configured to execute computer code stored in the volatile memory, and a graphical user interface coupled to the microprocessor At least one output interface of the computer system, the graphical user interface displaying a portion of the first digital color image next to the corresponding portion of the second digital color image by the user of the computer system One or more output interfaces, and one or more input interfaces coupled to a microprocessor, which enable visual comparison of the one or more input interfaces, the parts having significant changes, meaningful The part that did not change, Other receives a selection of a portion that requires further analysis may include one or more input interfaces, a.
For example, the present application provides the following items.
(Item 1)
A method for comparing a first digital color image and a second digital color image having aligned pixels, each pixel having data for N colors, the first digital color image comprising: , The appearance of a physical activity or object at a first time, the second digital color image has the appearance of a physical activity or object at a second time, and the method
Forming a composite image having pixels in a 2N dimensional color model, said forming comprising: color data for pixels in said first digital color image; and registration in said second digital color image Done by defining color data for each pixel in the composite image as a concatenation of color data for the pixel;
Determining a set of main components of the composite image using main component analysis, wherein the set of main components defines a coordinate system in the 2N dimensional color model;
Selecting a subset of main components in said coordinate system as main components each representing a difference rather than a correlation between said first digital color image and said second digital color image When,
Calculating, for each pixel in the composite image, a normalized pixel change value as a function of coordinates of pixel color data along each main component in the subset, the normalized pixel The change values collectively define a change image having a one-dimensional color model, and the calculated normalized pixel change values for each pixel are between the first and second time points. Measuring a change in said physical activity or object
Method, including.
(Item 2)
3. The method of claim 1, wherein N = 3, and each pixel comprises red, green and blue (RGB) color data.
(Item 3)
3. The method of claim 1, wherein N = 4 and each pixel has red, green, blue and infrared (RGB IR) color data.
(Item 4)
Selecting a subset of the main components may, for each main component of the composite image,
(A) by subtracting a first N-dimensional portion of the main component representing the first digital color image from a second N-dimensional portion of the main component representing the second digital color image , Forming an N-dimensional difference vector,
(B) selecting the main component for inclusion in the subset if and only if the N-dimensional length of the difference vector is greater than a given length;
The method according to item 1, including
(Item 5)
The method according to claim 1, wherein the normalized pixel change value for each pixel is calculated as the maximum of the coordinates of the color data of the pixel along each main component in the subset.
(Item 6)
The method according to claim 1, further comprising removing pixels having a brightness below a shadow threshold from the composite image before determining the set of main components of the composite image.
(Item 7)
The physical activity or object is a municipality, and the first digital color image represents the appearance of the municipality at a first time, and the second digital color image is a municipality at a second time. The municipality is divided into a plurality of compartments and the method
Dividing the change image into a plurality of segments: (a) each given segment in the plurality of segments is formed of a series of pixels, completely within the compartments in the plurality of compartments, (b 2.) The membership of each pixel in the given segment is determined as a function of its normalized pixel change value,
Calculating, for each segment in the plurality of segments, a segment change probability as a function of the normalized pixel change value of the pixels forming it;
Calculating, for each partition in the plurality of partitions, a partition change probability as a function of the segment change probability of the segment therein;
Ranking the compartments in the plurality of compartments according to their respective compartment change probabilities
The method of claim 1 further comprising
(Item 8)
11. The method of item 7, wherein the first digital color image and the second digital color image represent the appearance of a municipality approximately integer years apart.
(Item 9)
The method according to claim 7, wherein the segment change probability for each segment in the plurality of segments is calculated as an average of the normalized pixel change values of the pixels forming it.
(Item 10)
The method according to claim 7, further comprising excluding segments having a segment change probability less than a segment change threshold before calculating the partition change probability.
(Item 11)
The method according to claim 7, further comprising excluding a segment consisting of a number of pixels representing an area having a square foot number less than a segment size threshold before calculating the section change probability.
(Item 12)
The segment change probability for each segment in the plurality of segments is between the segment change probability of the given segment and the number of pixels in the given segment and the number of pixels in the segment containing the given segment 8. A method according to item 7, calculated as the sum of the product of the ratio and the ratio over each segment within a given compartment.
(Item 13)
Further comprising providing a graphical user interface (GUI) to a computer system, the GUI displays a portion of the first digital color image next to the corresponding portion of the second digital color image The method according to item 7, thereby enabling visual comparison of parts by the user of the computer system.
(Item 14)
In the GUI,
(A) a distinctive outline of a section within said portion of said first digital color image;
(B) salient contours of compartments in the corresponding portion of the second digital color image;
(C) a control to receive a selection indicating which compartments have changed significantly, which have not changed significantly, or which require further analysis
14. A method according to item 13, further comprising displaying simultaneously.
(Item 15)
14. The method according to item 13, further comprising providing a control that allows the user to adjust the shadow threshold, the segment change threshold, or the segment size threshold in the GUI.
(Item 16)
A non-volatile, tangible computer readable storage medium on which is stored computer program code for comparing a first digital color image having aligned pixels and a second digital color image. Each pixel has data for N colors, and the first digital color image has the appearance of physical activity or object at a first time, and the second digital color image is , The appearance of physical activity or objects at a second point in time, said storage medium being
Forming a composite image having pixels in a 2N dimensional color model, said forming comprising: color data for pixels in said first digital color image; and registration in said second digital color image Done by defining color data for each pixel in the composite image as a concatenation of color data for the pixel;
Determining a set of main components of the composite image using main component analysis, wherein the set of main components defines a coordinate system in the 2N dimensional color model;
Selecting a subset of main components in said coordinate system as main components each representing a difference rather than a correlation between said first digital color image and said second digital color image When,
Calculating, for each pixel in the composite image, a normalized pixel change value as a function of coordinates of pixel color data along each main component in the subset, the normalized pixel The change values collectively define a change image having a one-dimensional color model, and the calculated normalized pixel change values for each pixel are between the first and second time points. Measuring a change in said physical activity or object
A storage medium containing program code for performing.
(Item 17)
The program code for selecting a subset of the main components is for each main component of the composite image:
(A) by subtracting a first N-dimensional portion of the main component representing the first digital color image from a second N-dimensional portion of the main component representing the second digital color image , Forming an N-dimensional difference vector,
(B) selecting the main component for inclusion in the subset if and only if the N-dimensional length of the difference vector is greater than a given length;
18. A storage medium according to item 16, comprising program code for performing.
(Item 18)
The program code for calculating the normalized pixel change value for each pixel is a program code for selecting the maximum of the coordinates of the color data of the pixel along each main component in the subset. 20. A storage medium according to item 16 comprising:
(Item 19)
The program code for calculating the normalized pixel change value for each pixel comprises: program code for averaging the coordinates of the color data of the pixel along each main component in the subset 20. A storage medium according to item 16 comprising.
(Item 20)
The program code further includes program code for removing pixels having luminance less than a shadow threshold from the composite image before determining the set of main components of the composite image. Storage medium.
(Item 21)
The physical activity or object is a municipality, and the first digital color image represents the appearance of the municipality at a first time, and the second digital color image is a municipality at a second time. And the municipality is divided into a plurality of compartments, and the storage medium is
Dividing the change image into a plurality of segments: (a) each given segment in the plurality of segments is formed of a series of pixels, completely within the compartments in the plurality of compartments, (b 2.) The membership of each pixel in the given segment is determined as a function of its normalized pixel change value,
Calculating, for each segment in the plurality of segments, a segment change probability as a function of the normalized pixel change value of the pixels forming it;
Calculating, for each partition in the plurality of partitions, a partition change probability as a function of the segment change probability of the segment therein;
Ranking the compartments in the plurality of compartments according to their respective compartment change probabilities
20. A storage medium according to item 16, further comprising program code for performing.
(Item 22)
For each segment in the plurality of segments, the program code for calculating a segment change probability includes program code for calculating an average of the normalized pixel change values of the pixels forming it. The storage medium according to Item 21.
(Item 23)
22. A storage medium according to item 21, further comprising program code for eliminating a segment having a segment change probability less than a segment change threshold before calculating the partition change probability.
(Item 24)
22. A storage medium according to claim 21, further comprising program code for eliminating a segment consisting of a pixel number representing an area having a square foot number less than a segment size threshold before calculating the partition change probability.
(Item 25)
The program code for calculating the partition change probability for each partition in the plurality of partitions comprises the segment change probability of the given segment, the number of pixels in the given segment, and the given segment. 22. A storage medium according to item 21, comprising program code for calculating the sum of the product of the ratio between the number of pixels in the containing compartments over each segment in the given compartment.
(Item 26)
The computer program further comprises program code for providing a graphical user interface (GUI) to a computer system, the GUI displaying a portion of the first digital color image next to the corresponding portion of the second digital color image. 22. A storage medium according to item 21, which enables visual comparison of parts by the user of the computer system by displaying on.
(Item 27)
In the GUI,
(A) a distinctive outline of a section within said portion of said first digital color image;
(B) salient contours of compartments in the corresponding portion of the second digital color image;
(C) a control to receive a selection indicating which compartments have changed significantly, which have not changed significantly, or which require further analysis
26. A storage medium according to item 26, further comprising program code for displaying at the same time.
(Item 28)
The storage medium according to Item 26, further comprising program code for providing a control unit that enables a user to adjust a shadow threshold, a segment change threshold, or a segment size threshold in the GUI.
(Item 29)
A computer system for comparing a first digital color image and a second digital color image having aligned pixels, wherein each pixel comprises data for N colors, said first digital color image Has the appearance of physical activity or object at a first point in time, the second digital color image has the appearance of physical activity or object at a second point in time, and the computer system
Volatile memory storing computer program code therein, said computer program code comprising
Forming a composite image having pixels in a 2N dimensional color model, said forming comprising: color data for pixels in said first digital color image; and registration in said second digital color image Done by defining color data for each pixel in the composite image as a concatenation of color data for the pixel;
Determining a set of main components of the composite image using main component analysis, wherein the set of main components defines a coordinate system in the 2N dimensional color model;
Selecting a subset of main components in said coordinate system as main components each representing a difference rather than a correlation between said first digital color image and said second digital color image When,
Calculating, for each pixel in the composite image, a normalized pixel change value as a function of coordinates of pixel color data along each main component in the subset, the normalized pixel The change values collectively define a change image having a one-dimensional color model, and the calculated normalized pixel change values for each pixel are between the first and second time points. Measuring a change in said physical activity or object
Volatile memory, which is intended to
A microprocessor coupled to the volatile memory and configured to execute the computer code stored in the volatile memory;
One or more output interfaces coupled to the microprocessor for providing a graphical user interface, the graphical user interface including a portion of the first digital color image and the second digital color image One or more output interfaces that allow visual comparison of parts by the user of the computer system by displaying next to the corresponding parts of
One or more input interfaces coupled to the microprocessor, wherein the one or more input interfaces are displayed with significant change, displayed with no significant change, or further displayed. One or more input interfaces, with a selection that indicates the displayed part requesting analysis
Including computer systems.

  The features of the embodiments described above may be more readily understood by reference to the following detailed description in connection with the accompanying drawings.

FIG. 1 is a representation of a graphical user interface (GUI) for performing tax assessment in accordance with an embodiment of the present invention. FIG. 2 is a schematic representation of a computer system for performing tax assessment in accordance with an embodiment of the present invention. FIG. 3 is a flow chart illustrating a process for performing a tax assessment in accordance with an embodiment of the present invention. FIG. 4 is a representation of the GUI of FIG. 1 showing a list of compartments ranked according to their respective compartment change probabilities and the particular compartments selected from the list. FIG. 5 shows a GUI displaying a first partition with a high partition change probability prior to visual analysis by the analyst. FIG. 6 shows a GUI displaying a second partition with high partition change probability after the analyst has marked the first partition as significantly changed for the purpose of performing a tax assessment. FIG. 7 shows a GUI displaying a third partition with a high partition change probability after the analyst has marked the second partition as significantly changed for the purpose of performing a tax assessment. FIG. 8 shows a GUI displaying a fourth partition with low partition change probability prior to visual analysis. FIG. 9 shows a GUI displaying a fifth compartment with low compartment change probability after the analyst has marked the fourth compartment as not making significant changes for the purpose of performing a tax assessment. .

(Detailed Description of Specific Embodiments)
According to various embodiments of the present invention, systems, methods, and computer program products compare images of physical activities or objects. The pixel data from the first registered color image and the second registered color image defined in the N-dimensional color model are concatenated to form an image in the 2N-dimensional color model. Principal component analysis is performed on the composite image to identify components that represent changes rather than correlations. In the composite image, the coordinates for each pixel are determined along the main component that represents the change, and normalized pixel change values are calculated. Particularly in connection with performing a tax assessment, the pixels in the composite image are segmented according to their respective normalized pixel change values, and the segments do not mutually predefine geographic parcel boundaries. Partition change probabilities are calculated as a function of segment change probabilities (ie, a function of their normalized pixel change value). Finally, the partitions are ranked by their probability of change and highlighted in the graphical user interface.

  Definition. As used in the specification and the appended claims, the following terms may have the indicated meaning, unless the context requires otherwise.

  "Local government" refers to any political form that at least partially funds its operations by collecting taxes on real estate contained within its boundaries. "District" refers to real estate in a municipality assessed for taxable value to facilitate such funding.

  "Color space" refers to a specific composition of colors. "Color model" refers to the manner in which colors in color space may be represented as a tuple of numbers, each tuple representing a color or combination of colors.

  The "concatenation" of color data adds the second vector of tuples to the first vector of tuples, which is equal to the sum of the length of the first vector and the length of the second vector It means creating a third vector of tuples with length.

  "Main component analysis" of data in vector space is a process for determining the orthogonal transformation that rotates the coordinate system of vector space, so that the maximum variance (width) of the data is The second largest dispersion of data occurs along the second axis of rotation (referred to as the second "main component"), which occurs along the first axis of rotation (referred to as the first "main component"). Etc., means a process. Principal Component Analysis has various different names in different fields: among many other names and applications, in mathematical engineering this is an eigen-orthogonal decomposition, in linear algebra it is Singular value decomposition or eigen value decomposition of a matrix, and in signal processing this is essentially a transformation of the Karhunen-Loeve theorem, in statistics this is the canonical (auto) correlation of a single data set It is an analysis.

  "Segmentation" of a digital image refers to a process for partitioning the pixels of the image into successive groups, each such group of pixels having a common desired property (but not limited to). , Color, intensity, or texture etc.).

  FIG. 1 is a representation of a graphical user interface (GUI) 100 for performing tax assessment, in accordance with one embodiment of the present invention. The GUI 100 is used to compare the “before” image 110 and the “after” image 120 for changes to compartments that may affect the tax assessment value. One such significant change is the newly paved area 130 at the back of the section (when viewed from the street). Another is shown by the closed private road 140, which indicates that the construction is complete at a particular house. Yet another is an add-on 150 on the home. An example of a change that would not affect tax assessment is illustrated by the new landscaping 160. Although the GUI 100 of FIG. 1 only shows a small number of divisions in a municipality, for example, software that implements a GUI or the like is a control unit for enabling a user to display all or a part of the municipality. It should be understood that can be included.

  FIG. 2 is a schematic representation of a computer system 200 for performing tax assessment in accordance with an embodiment of the present invention. While computer system 200 of FIG. 2 is conventional, one skilled in the art will appreciate that other types of computer systems may be used in accordance with various embodiments of the present invention.

  Computer system 200 includes a central processing unit (CPU) 210 that performs arithmetic and logic calculations, such as, for example, a microprocessor known in the art. Computer system 200 also includes volatile memory 220 for temporarily storing data, such as, for example, random access memory (RAM), as known in the art. Computer system 200 includes non-volatile memory 230 for persistently storing data, such as, for example, hard disk drives known in the art. Computer system 200 further includes a peripheral interface. Output interface 240 is connected to output device 242 to provide information from computer system 200 to the outside world. Output devices 242 may include, for example, video displays and speakers, and each output interface 240 may be a video socket on a graphics processing unit (GPU) and an audio jack on a sound card. Input interface 250 is coupled to input device 252 for providing information from the outside world into computer system 200. Input devices 252 may include, for example, a keyboard and a computer mouse, and each input interface 250 may be a serial, parallel, or universal serial bus (USB) port. Internally, the functionality of output interface 240 and input interface 250 may be provided by one or more integrated circuits, or by CPU 210.

  FIG. 3 is a flow chart illustrating a process for performing a tax assessment in accordance with an embodiment of the present invention. Although the tax valuation method of FIG. 3 includes eight processes, variations of the methods may include additional processes or optional processes, as described below, or perform these processes in a different order obtain. In an exemplary embodiment of the invention, the method is performed by hardware, software, or a combination thereof internal to computer system 200. For the sake of particularity, the process of FIG. 3 is described herein in the context of performing a tax assessment, but the inventive concept is to determine a significant change between two images It should be understood by one skilled in the art that it may be applied in other contexts as desired.

  In the preferred embodiment of the present invention, two digital color images are loaded into the computer system prior to starting the process of FIG. In particular, the first digital color image represents a "front" image for comparison, while the second digital color image represents a "back" image for comparison. The images represent the appearance of the municipality at different times. Thus, the first image may represent the municipality at a first point in time when the tax has been assessed, while the second image may represent the municipality at a point in time when the tax has not been assessed. In particular, the first digital color image and the second digital color image may represent the appearance of a municipality approximately an integer number of years apart if the municipality performs tax assessments in certain months of a year not necessarily every year. The images are aligned (aligned), for example as shown in FIG.

  Each digital image uses a color model that includes N colors. Thus, if each pixel can be described by three colors (eg, red, green and blue), then N = 3 for the color model, and each pixel is a three-dimensional vector whose coordinates are RGB color values It can be considered to be describing. If infrared data is also included in each pixel, then N = 4 and each pixel describes a four dimensional vector whose coordinates are RGB IR color values. It should be understood that although three or four colors are typical, the inventive concept can be applied to any number of colors, ie the color model has an arbitrary value N.

  In process 301, the computer system forms a composite image from the first digital color image and the second digital color image. The composite image has twice the color dimension, and if each of the first and second images has a color dimension N, the composite image has a color dimension 2N. Process 301 may be performed by concatenating color data for each pixel in the first image and color data for the corresponding aligned pixel in the second image to form a 2N-dimensional vector. Taking the origin as a starting point, each such vector defines a data point in a 2N dimensional color model. Taken together, a set of pixels in the composite image form a set of such data points in a 2N dimensional space. Each such data point simultaneously encodes the color of two registered pixels (one from the first image and one from the second image).

  At this stage, the optimization optionally includes compensation for shadows. Pixels in the shadow may be detected when the brightness or intensity of the pixels is low. As each pair of images may have different light and dark conditions, it is advantageous to provide a shadow threshold (below which pixels are excluded from further processing). Furthermore, once it is determined that a pixel in one digital color image is in the shadow, the corresponding aligned pixel may be discarded (and thereby the corresponding pixel in the composite image). Thus, the method optionally includes removing, from the composite image, pixels having a brightness or intensity below the shadow threshold.

  In process 302, the computer system uses main component analysis or PCA (this phrase is known in the art) to determine the set of main components of the data points of the remaining pixels in the composite image. PCA can be thought of as finding a preferred direction in data that describes how wide the data is. Each such preferred direction corresponds to some feature of the data. The output of process 302 is a new rotated coordinate system in a 2N-dimensional color model, whose axes do not necessarily represent colors, but rather represent linear combinations of colors that characterize the digital image. There is. Because many algorithms for performing PCA are known in the art, and the choice of algorithm can affect the performance of calculations, embodiments of the present invention are most likely given the hardware constraints of the computer system. Any efficient algorithm may be used.

  In the context of the present invention, the first main component in the 2N-dimensional color model usually corresponds to the overall scene brightness, which is the correlation between the first digital image and the second digital image. To understand this, the two images are most easily visually compared if they were taken from the same angle at the same time of day, and the corresponding registered pixels are mutually It should be noted that the brightness is almost the same. Thus, the correlation in luminance provides a strong signal in the 2N dimensional color model. Similarly, the second main component usually corresponds to the difference in scene brightness between the two images. In contrast to the first main component, the second main component in this case represents the difference, rather than the correlation, between the first digital color image and the second digital color image, Therefore, it suggests a significant change between the images.

  In many cases, the third main component represents the correlation between pixel colors. As expected, the aligned pixels typically represent the same land cover, and the color of most of the land cover will not differ between images. Thus, a strong correlation will be present in the composite image, but perhaps not as strong as the correlation between overall scene brightness. Similarly, the fourth main component may represent the difference between the colors of the registered pixels, and thus may make a meaningful change between the images.

  It should be understood that this description of the main components is exemplary only, and each comparison of digital images using process 302 may generate main components corresponding to other features of the image. The main components representing the difference between the two images rather than the correlation are not necessarily the second main component and the fourth main component, but rather the main components of the difference are different subsets It should be understood that it may be. In any event, as can be seen from the above discussion, some of the main components represent the correlation between the first digital color image and the second digital color image, while the other main components are Elements represent differences.

  In process 303, the computer system selects the main components in the rotated coordinate system, each main component rather than the correlation between the first digital color image and the second digital color image , Represents the difference. A human being can easily determine which main component represents a difference, but the computer system has to rely on trial and error.

  Here, a two-step procedure for selecting these is described. The procedure works on a 2N × 2N linear transformation matrix that rotates the original 2N color band into the coordinate system defined by the 2N main components by matrix multiplication instead of working on the main components themselves Do. The transformation matrix has coefficients or weights that represent how much each original color band is present in the corresponding main component.

  First, form a first N-dimensional portion (vector) of the coefficients of the transformation matrix corresponding to the first image, and a second N-dimensional portion (vector) of the transformation matrix corresponding to the second image Form This step involves separating the 2N coefficients of the columns of the transformation matrix into 2N-dimensional vectors, since the concatenation in process 301 formed the color values of the change image. The first vector and the second vector represent parts of a linear transformation and do not correspond directly to any color data present in any pixel of any of the original image Should be understood.

  Next, a difference vector is obtained by subtracting the first vector from the second vector. If the first vector and the second vector have similar magnitudes and coordinates of the sign, the difference vector may have small magnitudes indicative of correlation, while the first vector and the second If the vectors have coordinates whose signs are opposite, the difference vector may have a large magnitude that indicates a change. Therefore, the N-dimensional length of the difference vector is a measure of how much the main component represents a change. In fact, comparison of the length of the difference vector to a given length provides a binary test that selects a subset of the main components representing the difference rather than the correlation. The given length may be, for example, length 1 if the color values are to be scaled between 0 and 1. It should be understood that other lengths may be used and other procedures for implementing process 303 may be used.

  It is worth noting that at this point the number of main components selected by process 303 is less than the total number (2N) of main components. Since each main component potentially represents a significant change, a digital color image with more color bands (i.e. a higher value of N) is potentially a change It will generate more opportunities for detecting. This observation follows from the fact that color data containing more color data contains more correlations and more differences between these data. However, valid results can be obtained using uniform full-color matching images where N = 1, in such images the first main component represents a rough correlation between the images, The main components of 2 represent the coarse differences between the images.

  Next, process 304 calculates, for each pixel in the composite image, normalized pixel change values as a function of the coordinates of the pixel color data along each main component in the subset generated by process 303. . This process 304 takes the maximum value of the pixel's coordinates along the selected subset (which represents the difference), while ignoring the pixel's coordinates along the other main components (which , Representing correlation) can be implemented.

  The normalized pixel change values collectively define a change image having a one-dimensional color model. This change image may be visualized as a grayscale image that represents a significant change between the first color image and the second color image. That is, the results of processes 301-304 first map each pixel in the 2N-dimensional composite image into a lower-dimensional subspace and then into a single (one-dimensional) normalized pixel change value. It will be mapping. Each pixel in the composite image shares a spatial relationship with all other pixels defined by the alignment between the first color image and the second color image, this relationship being a normalized pixel By being applied to the change values, a change image having a single "color" dimension can be generated.

  After performing steps 301-304, useful information about the significant change between the first digital color data and the second digital color data is already obtained in the form of a gray scale change image It should be understood. This registration of the variance image with respect to the first and second images provides an advantageous method for comparing the two original images. In particular, the variation image having high "intensity" represents a large variation between the respective registered images in the first digital color image and the second digital color image.

  However, for the purpose of illustrating the particular application of this variation image, the method of FIG. 3 follows four additional processes that are useful for performing tax assessment. Thus, for the remainder of the description of FIG. 3, the physical object represented by the image is a municipality, and the first digital color image is at a first point in time (e.g. around the point in time of previous tax assessment) The second digital color image may represent the appearance of the municipality at a second point in time (e.g., a recent point in time), and it may be assumed that the municipality is divided into a plurality of compartments.

  At step 305, the computer system divides the change image into several segments, each containing pixels that share common visual characteristics. Ideally, the common property relates to the type of land cover (e.g. grass or roof) and each segment represents a pixel imaging a continuous area having the same type of land cover. Many algorithms for segmenting digital images are known in the art, such as formation thresholds, clustering, compression, formation histograms, detected edges, growth regions, partitioning graphs, and the like. It should be understood that the scope of the present invention is not necessarily limited by the choice of the segmentation algorithm. In the illustrated embodiment, the membership of each pixel in the segment is determined as a function of its normalized pixel change value. Thus, for example, two neighboring pixels may be in the same segment if their normalized pixel change values are within a certain percentage of one another.

  Each segment is formed of consecutive pixels. Because the segments are continuous, each defines a shape that can be advantageously visualized in a graphical user interface using vector graphics. Each segment may be completely in a single compartment in multiple compartments, but different segments may be in different compartments. Thus, the segments formed by process 305 can be used in subsequent calculations to determine the overall partition change probability.

  In process 306, the computer system calculates, for each segment, a segment change probability as a function of the normalized pixel change value of the pixels forming it. This segment change probability can be seen as a "zone average" of the change values of the pixels that make up the segment. By performing this zonal averaging, each segment simply represents a normalized measure of the change between the first digital image and the second digital image, segment by segment rather than pixel by pixel. A number is given.

  At this stage, one optimization optionally involves excluding segments that have a segment change probability less than the segment change threshold. Thus, the computer system may advantageously be instructed that particular segments representing insignificant changes should be excluded from further consideration. Such instructions may take the form of controls on a graphical user interface such as, for example, a slider bar.

  Further optimization optionally includes excluding segments consisting of pixel counts representing areas having less than a given square foot according to a segment size threshold. Thus, advantageously, the computer system may be instructed that certain small segments should be excluded from further consideration, even if they represent a significant change. Such instructions may also take the form of controls in a graphical user interface.

  In process 307, the computer system calculates, for each partition, the partition change probability as a function of the segment change probability of the segments within it. At process 306, as the pixel change values are averaged to generate segment change values, process 307 averages the segment change values to generate a segment change value. However, while each pixel in the composite image represents the same amount of geographic area, the segments may vary in size such that small adjustments may be made to the averaging process.

  Thus, in one embodiment of the present invention, the partition change probabilities for each partition are calculated as the sum of the segment change probabilities weighted by their respective areas. That is, the partition change probability is a given partition of the product of the segment change probability of a given segment and the ratio between the number of pixels in a given segment and the number of pixels in a partition containing a given segment It is the sum over each segment within. It should be understood that other methods for calculating the parcel change probability may be employed without departing from the inventive concept.

  Finally, in process 308, the computer system ranks the parcels according to their respective parcel change probabilities, thereby providing the tax evaluator with a suggested order for visiting the parcels to perform the tax assessment. provide. Although this ranking may be performed using conventional sorting means, it should be understood that the scope of the present invention is not limited by the choice of sorting algorithm. Any such algorithm can be chosen to optimize the efficiency of the ranking process.

  Some embodiments of the present invention provide a graphical user interface (GUI) to a computer system. A screen shot of such a GUI is provided in FIG. 1, and another is provided in FIG. 4 described below. The GUI can display a portion of the first digital color image next to the corresponding aligned portion of the second digital color image, thereby allowing the tax evaluator using the computer system to view the portion It may be possible to make a comparison. In particular, the GUI may facilitate the comparison by being able to simultaneously display the salient contours of the compartments defined by the aligned portions of the two images. Also, the GUI may include controls that allow the tax rater to adjust the shadow threshold, segment change probability, threshold, or segment size threshold.

  FIG. 4 is a representation of the graphical user interface 100 of FIG. 1 and shows a list of compartments ranked according to their respective compartment change probabilities and the particular compartments selected from the list. Some of the features described above can be seen in FIG. In particular, the first digital color image 410 and the second registered digital color image 420 can be viewed. Images 410 and 420 represent zoomed displays of the “front” image 110 and the “back” image 120, respectively, to allow detailed investigation of a small number of segments. The given section is prominently contoured in a portion 430a of the first digital color image and in a corresponding portion 430b of the second digital color image.

  Three sensitivity controls are provided in the embodiment of FIG. The change sensitivity control unit 441 enables the tax evaluator or other users to specify the threshold of the segment change probability. The detail sensitivity control unit 442 allows the user to specify a segment size threshold. The shadow sensitivity control unit 443 enables the user to specify a shadow threshold.

  A list 450 of compartments ranked by their likelihood of change is also visible in the GUI. Sections are identified using various descriptive information such as, for example, their geographic coordinates, unique section identification numbers, or other such data that may be selectable using the display label control unit 451. . Displayed next to each identifier in this embodiment are a status indicator (the function of which will be described in connection with FIGS. 5-9) and a "change" bar (the length of which is shown in FIG. 3). And visually indicate the partition change probability calculated for the partition using the method described in the related. The data in each column may be stored according to the user's preferences. Thus, the partitions may advantageously be stored in descending order of their respective partition change probabilities, as shown in FIG. Thus, the user can be advanced to consider the partitions in the order suggested by this sorting.

  5-7 show the three sections with the highest probability of change. In particular, FIG. 5 shows a GUI displaying a first partition having a high partition change probability prior to visual analysis. The status of each partition is unknown, as indicated by the question mark in the "Status" column of the ranked partition list. The analyst performs a visual comparison of the indicated sections only, and manipulates one of several controls based on the analysis. Thus, if the analyst sees that the partition has undergone a significant change for the purpose of performing a tax assessment, the analyst may operate the “mark change” control 501. Or, if it appears that the section has not changed significantly, the analyst may operate the “mark invariant” control 502. If further analysis is required, the analyst may operate the “mark needed review” control 503. In this case, the analyst can operate the first control unit 501 because the section is clearly improved. According to a particular efficient embodiment of the invention, the operation of any of the controls 501, 502, 503 advantageously advances the visualization to the next compartment.

  FIG. 6 shows a GUI displaying a second partition having a high partition change probability after the analyst has marked the first partition as significantly changed. According to one embodiment of the present invention, marking the first partition causes a "changed" triangle to be represented in the "state" column 601 of the list of ranked partitions. This triangle indicates that the compartment has been analyzed and determined to have changed. In some embodiments, the triangle may be colored in a suggestive color (eg, green) to indicate that a positive match has been determined by the analyst. FIG. 7 shows a GUI displaying a third partition having a high partition change probability after the analyst has marked the second partition as significantly changed. Here, it should be noted that the "state" column 601 has two triangles.

  FIG. 8 shows a GUI displaying a fourth partition having a low partition change probability prior to visual analysis. How much shorter the "variation" bar is for the compartments shown compared to those for the compartments shown in FIGS. 5-7 is that the probability that the compartments have significantly changed between the two digital images is much more Note that it stands for low. The tax analyst may consider that the visual change does not represent any significant change, and may operate a "mark invariant" control. The results are shown in FIG. 9, which displays the fifth section with low section change probability after the analyst has marked the fourth section as not having made a significant change. Shows a GUI.

  Note the appearance of the square in the state column 601 for the fourth section marked "invariant" here in FIG. This square indicates that the section was analyzed and determined to have not changed. In some embodiments, squares may be colored in a suggestive color (eg, red) to indicate that the analyst determined a negative match.

  It should be understood that various embodiments of the present invention may include other features such as, for example, the ability to record analysts' decisions in files interoperable with other computer systems and programs. One skilled in the art can still appreciate other modifications to the computer system or GUI using the inventive concepts described herein.

  The embodiments of the present invention described above are intended to be exemplary only, and many variations and modifications may be understood by those skilled in the art. All such variations and modifications are intended to be within the scope of the present invention as defined in any of the appended claims.

  It should be noted that the logical flow diagram is used herein to illustrate the various aspects of the invention, and to limit the invention to any particular logical flow or implementation. It should not be conceived. The logic described may be party to different logic blocks (e.g. programs, modules, functions or subroutines) without changing the overall result or without departing from the true scope of the present invention. Can be Often, logical elements can be added, modified, omitted, or performed in a different order without changing the overall result or deviating from the true scope of the present invention. Or may be implemented using different logical configurations (eg, logical gates, looping primitives, state logic, and other logical configurations).

  The invention may be embodied in many different forms, such as computer program logic (eg, a microprocessor, microcontroller, digital signal processor, or general purpose computer, etc.) for use with a processor, programmable logic Includes programmable logic (eg, field programmable gate array (FPGA) or other PLD), discrete components, integrated circuit (eg, application specific integrated circuit (ASIC)), or any combination thereof for use with the device Although any other means etc. are included, it is not limited to these by any means.

  Computer program logic that implements all or part of the functionality already described in this specification may be embodied in various forms, such as source code forms, computer executable forms, and various other forms. Intermediate forms (e.g., forms generated by an assembler, a compiler, or a locator, etc.), but are by no means limited thereto. The source code is for use with various operating systems or environments, such as object code, assembly language, or a high level language such as, for example, Fortran, C, C ++, JAVA, or HTML. Etc.) may be included. The source code may define and use various data structures and communication messages. The source code may be in computer executable form (eg, via an interpreter), or the source code may be converted to computer executable form (eg, via a translator, assembler, or compiler) .

  The computer program may be tangible storage medium (eg, semiconductor memory device (eg, RAM, ROM, PROM, EEPROM, or flash programmable RAM), magnetic memory device (eg, diskette or fixed disk), optical memory device (eg, CD) In any form (eg source code form, computer executable form or intermediate form), either permanent or temporary, in a ROM), PC card (eg PCMCIA card) or other memory device etc) ) Can be fixed. The computer program may be any of a variety of communication technologies (including but not limited to analog technology, digital technology, optical technology, wireless technology (eg Bluetooth), networking technology, and interworking technology) Either can be fixed in any form within the signal that can be transmitted to the computer. The computer program may be distributed in any form, such as a removable storage medium (e.g. shrink-wrapped software) with an accompanying printed or electronic document, and may be preloaded on the computer system (e.g. system ROM or fixed disk) Or from a server or electronic bulletin board via a communication system (eg, the Internet or the World Wide Web).

  Whether hardware logic (including programmable logic for use with programmable logic devices) implementing all or part of the functionality already described herein may be designed using traditional manual methods, Alternatively, it can be designed electronically using various tools such as, for example, computer aided design (CAD), hardware description language (eg VHDL or AHDL), or PLD programming language (eg PALASM, ABEL or CUPL) , Can be captured, simulated, or documented.

  Programmable logic includes tangible storage media (eg, semiconductor devices (eg, RAM, ROM, PROM, EEPROM, or flash programmable RAM), magnetic memory devices (eg, diskette or fixed disk), optical memory devices (eg, CD) -In ROM) or other memory devices etc.) may be fixed either permanently or temporarily. Programmable logic may include, but is not limited to, various communication technologies (including but not limited to analog technology, digital technology, optical technology, wireless technology (eg, Bluetooth), networking technology, and interworking technology). Either can be fixed in the signal that can be transmitted to the computer. The programmable logic may be distributed as a removable storage medium (e.g., shrink wrap software) with associated printed or electronic documents and may be preloaded on a computer system (e.g., on a system ROM or fixed disk) or a communication system (e.g. For example, it can be distributed from a server or an electronic bulletin board via the Internet or the World Wide Web.

Claims (27)

A method for comparing a first digital color image and a second digital color image having aligned pixels, each pixel having data for N colors, the first digital color image comprising: , The appearance of a physical activity or object at a first time, the second digital color image has the appearance of a physical activity or object at a second time, and the method
Forming a composite image having pixels in a 2N dimensional color model, said forming comprising: color data for pixels in said first digital color image; and registration in said second digital color image Done by defining color data for each pixel in the composite image as a concatenation of color data for the pixel;
Determining a set of main components of the composite image using main component analysis, wherein the set of main components defines a coordinate system in the 2N dimensional color model;
Selecting a subset of main components in said coordinate system as main components each representing a difference rather than a correlation between said first digital color image and said second digital color image When,
Calculating, for each pixel in the composite image, a normalized pixel change value as a function of coordinates of pixel color data along each main component in the subset, the normalized pixel The change values collectively define a change image having a one-dimensional color model, and the calculated normalized pixel change values for each pixel are between the first and second time points. Measuring a change in said physical activity or object.
Selecting a subset of the main components may, for each main component of the composite image,
(A) by subtracting a first N-dimensional portion of the main component representing the first digital color image from a second N-dimensional portion of the main component representing the second digital color image , Forming an N-dimensional difference vector,
(B) selecting the main component for inclusion in the subset if and only if the N-dimensional length of the difference vector is greater than a given length, .   The method of claim 1, wherein N = 3, and each pixel comprises red, green and blue (RGB) color data.   The method according to claim 1, wherein N = 4 and each pixel comprises red, green, blue and infrared (RGB IR) color data.   The method of claim 1, wherein the normalized pixel change value for each pixel is calculated as the maximum of the coordinates of the color data of the pixel along each main component in the subset.   The method according to claim 1, further comprising removing pixels having a brightness below a shadow threshold from the composite image before determining the set of main components of the composite image. The physical activity or object is a municipality, and the first digital color image represents the appearance of the municipality at a first time, and the second digital color image is a municipality at a second time. The municipality is divided into a plurality of compartments and the method
Dividing the change image into a plurality of segments: (a) each given segment in the plurality of segments is formed of a series of pixels, completely within the compartments in the plurality of compartments, (b 2.) The membership of each pixel in the given segment is determined as a function of its normalized pixel change value,
Calculating, for each segment in the plurality of segments, a segment change probability as a function of the normalized pixel change value of the pixels forming it;
Calculating, for each partition in the plurality of partitions, a partition change probability as a function of the segment change probability of the segment therein;
The method of claim 1, further comprising: ranking the compartments in the plurality of compartments according to their respective compartment change probabilities.   7. The method of claim 6, wherein the first digital color image and the second digital color image represent the appearance of a municipality approximately integer years apart.   7. The method of claim 6, wherein the segment change probability for each segment in the plurality of segments is calculated as an average of the normalized pixel change values of the pixels forming it.   7. The method of claim 6, further comprising excluding segments having a segment change probability less than a segment change threshold prior to calculating the partition change probability. 7. The method of claim 6, further comprising excluding a segment consisting of a number of pixels representing an area having a square foot less than a segment size threshold prior to calculating the partition change probability.   The segment change probability for each segment in the plurality of segments is between the segment change probability of the given segment and the number of pixels in the given segment and the number of pixels in the segment containing the given segment 7. A method according to claim 6, calculated as the sum of the product of the ratio of and over each segment within a given compartment.   Further comprising providing a graphical user interface (GUI) to a computer system, the GUI displays a portion of the first digital color image next to the corresponding portion of the second digital color image 7. The method of claim 6, thereby enabling visual comparison of portions of the computer system by a user. In the GUI,
(A) a distinctive outline of a section within said portion of said first digital color image;
(B) salient contours of compartments in the corresponding portion of the second digital color image;
13. The method according to claim 12, further comprising: simultaneously displaying (c) significantly changed compartments, not significantly changed compartments, or a control for receiving a selection indicating compartments requiring further analysis. the method of.   The method according to claim 12, further comprising providing a control that allows a user to adjust a shadow threshold, a segment change threshold, or a segment size threshold in the GUI. A non-volatile, tangible computer readable storage medium on which is stored computer program code for comparing a first digital color image having aligned pixels and a second digital color image. Each pixel has data for N colors, and the first digital color image has the appearance of physical activity or object at a first time, and the second digital color image is , The appearance of physical activity or objects at a second point in time, said storage medium being
Forming a composite image having pixels in a 2N dimensional color model, said forming comprising: color data for pixels in said first digital color image; and registration in said second digital color image Done by defining color data for each pixel in the composite image as a concatenation of color data for the pixel;
Determining a set of main components of the composite image using main component analysis, wherein the set of main components defines a coordinate system in the 2N dimensional color model;
Selecting a subset of main components in said coordinate system as main components each representing a difference rather than a correlation between said first digital color image and said second digital color image When,
Calculating, for each pixel in the composite image, a normalized pixel change value as a function of coordinates of pixel color data along each main component in the subset, the normalized pixel The change values collectively define a change image having a one-dimensional color model, and the calculated normalized pixel change values for each pixel are between the first and second time points. Measuring the change in the physical activity or the object.
Selecting a subset of the main components may, for each main component of the composite image,
(A) by subtracting a first N-dimensional portion of the main component representing the first digital color image from a second N-dimensional portion of the main component representing the second digital color image , Forming an N-dimensional difference vector,
And (b) selecting the main component for inclusion in the subset if and only if the N-dimensional length of the difference vector is greater than a given length, Medium.   The program code for calculating the normalized pixel change value for each pixel is a program code for selecting the maximum of the coordinates of the color data of the pixel along each main component in the subset. A storage medium according to claim 15, comprising   The program code for calculating the normalized pixel change value for each pixel comprises: program code for averaging the coordinates of the color data of the pixel along each main component in the subset A storage medium according to claim 15, comprising. The program code according to claim 15, wherein the program code further includes program code for removing a pixel having a luminance below a shadow threshold from the composite image before determining the set of main components of the composite image. Storage medium. The physical activity or object is a municipality, and the first digital color image represents the appearance of the municipality at a first time, and the second digital color image is a municipality at a second time. And the municipality is divided into a plurality of compartments,
The storage medium is
Dividing the change image into a plurality of segments: (a) each given segment in the plurality of segments is formed of a series of pixels, completely within the compartments in the plurality of compartments, (b 2.) The membership of each pixel in the given segment is determined as a function of its normalized pixel change value,
Calculating, for each segment in the plurality of segments, a segment change probability as a function of the normalized pixel change value of the pixels forming it;
Calculating, for each partition in the plurality of partitions, a partition change probability as a function of the segment change probability of the segment therein;
The storage medium of claim 15, further comprising: program code for: ranking partitions in the plurality of partitions according to their respective partition change probabilities.   For each segment in the plurality of segments, the program code for calculating a segment change probability includes program code for calculating an average of the normalized pixel change values of the pixels forming it. 20. A storage medium according to claim 19.   20. The storage medium of claim 19, further comprising program code for eliminating segments having a segment change probability less than a segment change threshold prior to calculating the partition change probability.   20. The storage medium of claim 19, further comprising program code for eliminating a segment consisting of a number of pixels representing an area having a square foot number less than a segment size threshold prior to calculating the partition change probability.   The program code for calculating the partition change probability for each partition in the plurality of partitions comprises the segment change probability of the given segment, the number of pixels in the given segment, and the given segment. 20. A storage medium according to claim 19, comprising program code for calculating the sum of the product of the ratio between the number of pixels in the containing compartments over each segment in the given compartment.   The computer program further comprises program code for providing a graphical user interface (GUI) to a computer system, the GUI displaying a portion of the first digital color image next to the corresponding portion of the second digital color image. 20. A storage medium according to claim 19, wherein displaying on allows a visual comparison of parts by the user of the computer system. In the GUI,
(A) a distinctive outline of a section within said portion of said first digital color image;
(B) salient contours of compartments in the corresponding portion of the second digital color image;
(C) the program code for simultaneously displaying the significantly changed compartments, the non-significantly changed compartments, or the control unit for receiving a selection indicating a compartment requiring further analysis. The storage medium of 24.   The storage medium according to claim 24, further comprising program code for providing a control unit that enables a user to adjust a shadow threshold, a segment change threshold, or a segment size threshold in the GUI. A computer system for comparing a first digital color image and a second digital color image having aligned pixels, wherein each pixel comprises data for N colors, said first digital color image Has the appearance of physical activity or object at a first point in time, the second digital color image has the appearance of physical activity or object at a second point in time, and the computer system
Volatile memory storing computer program code therein, said computer program code comprising
Forming a composite image having pixels in a 2N dimensional color model, said forming comprising: color data for pixels in said first digital color image; and registration in said second digital color image Done by defining color data for each pixel in the composite image as a concatenation of color data for the pixel;
Determining a set of main components of the composite image using main component analysis, wherein the set of main components defines a coordinate system in the 2N dimensional color model;
Selecting a subset of main components in said coordinate system as main components each representing a difference rather than a correlation between said first digital color image and said second digital color image When,
Calculating, for each pixel in the composite image, a normalized pixel change value as a function of coordinates of pixel color data along each main component in the subset, the normalized pixel The change values collectively define a change image having a one-dimensional color model, and the calculated normalized pixel change values for each pixel are between the first and second time points. Measuring the change in the physical activity or the object,
Selecting a subset of the main components may, for each main component of the composite image,
(A) by subtracting a first N-dimensional portion of the main component representing the first digital color image from a second N-dimensional portion of the main component representing the second digital color image , Forming an N-dimensional difference vector,
(B) selecting the main component for inclusion in the subset if and only if the N-dimensional length of the difference vector is greater than a given length, volatilization Memory and
A microprocessor coupled to the volatile memory and configured to execute the computer code stored in the volatile memory;
One or more output interfaces coupled to the microprocessor for providing a graphical user interface, the graphical user interface including a portion of the first digital color image and the second digital color image One or more output interfaces that allow visual comparison of parts by the user of the computer system by displaying next to the corresponding parts of
One or more input interfaces coupled to the microprocessor, wherein the one or more input interfaces are displayed with significant change, displayed with no significant change, or further displayed. A computer system, comprising: one or more input interfaces that receive a selection indicating a displayed portion requesting analysis.