JP2017073071A - Identity determination program, identity determination device, and identity determination method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce an erroneous determination as to whether subjects are identical with each other.SOLUTION: It is determined whether pixels in predetermined areas are color-mixed with a background in a first image and a second image, respectively and when color mixture is determined, the color-mixed pixels are deleted from the predetermined areas and feature quantity in the predetermined areas where the color-mixed pixels are deleted is calculated. On the basis of the feature quantity in the predetermined areas of the respective first and second images, it is determined whether or not a subject in the predetermined area of the first image is identical with a subject in the predetermined area of the second image.SELECTED DRAWING: Figure 2

Description

  The present invention relates to an identical determination program, an identical determination device, and an identical determination method.

  Person movement information is used to eliminate congestion and improve services by optimizing stores and personnel in commercial facilities. By detecting a person from the video of the surveillance camera and associating a person who moves across multiple cameras, it is possible to automatically grasp the movement of a wide range of persons.

  Conventionally, characteristics such as the color and pattern of a person's clothes are used to determine whether or not they are the same person. At this time, in order to obtain the same feature from the same person in different surveillance cameras, only the person area is extracted so as not to be affected by the background, and it is determined whether or not the person is the same based on the characteristics of the person area. doing.

  An apparatus is known that extracts an arbitrary moving object from an image, changes or adjusts the color, and combines it with another image (see, for example, Patent Document 1). An apparatus for detecting the contour of an object in an image and extracting the object is known (see, for example, Patent Document 2). An apparatus is known that extracts an arbitrary object in an image, changes the color or shape of the extracted image, overlays the original image or background image, and creates a new image (for example, Patent Document 3). reference).

  Also, a moving object detection method described in Non-Patent Document 1 and a human detection method described in Non-Patent Document 2 are known.

JP-A-5-120417 JP-A-6-203157 JP-A-7-203299

Shigeki Nagaya, 4 others, “Detection of moving objects using time-correlated background judgment method”, IEICE Transactions, D-II, vol.79, no.4, pp.568-576, 1995 Atsushi Yamauchi and two others, “Human Detection Based on Statistical Learning from Images”, IEICE Transactions, vol.J96-D, No.9, pp.2017-2040, 2013

  In an image taken with a surveillance camera, the outline pixel of a person may be mixed with the color around the person (background). When the color mixture of the contour pixels occurs, there is a high possibility of erroneous determination as to whether or not they are the same person.

  In one aspect, an object of the present invention is to reduce the possibility of erroneous determination when determining whether or not subjects included in a plurality of images are the same.

  The same determination program according to the embodiment determines whether the first pixel in the first area in the first image is mixed with the first background pixel adjacent to the first area in the computer, and When it is determined that one pixel is mixed with the first background pixel, the first feature amount of the first region excluding the first contour pixel is calculated.

  In the second image, it is determined whether the second pixel in the second area is mixed with the second background pixel adjacent to the second area, and the second pixel is mixed with the second background pixel. If it is determined, the second feature amount of the second region excluding the second contour pixel is calculated.

  Based on the first feature quantity and the second feature quantity, it is determined whether or not the subject in the first area and the subject in the second area are the same.

  According to the same determination program according to the embodiment, it is possible to reduce the possibility of erroneous determination when determining whether or not subjects included in a plurality of images are the same.

It is a figure which shows a part of image when color mixing generate | occur | produces. It is a lineblock diagram of a person judging system concerning an embodiment. It is an example of a feature information table. It is a flowchart of the person area extraction process which concerns on embodiment. It is a figure which shows an attention pixel, an inner side pixel, and an outer side pixel. It is a figure which shows the position of the inner side and outer side pixel with respect to an attention pixel. It is a figure which shows the position of the inner side and outer side pixel with respect to an attention pixel. It is a figure which shows the position of the inner side and outer side pixel with respect to an attention pixel. It is a figure which shows the position of the inner side and outer side pixel with respect to an attention pixel. It is a figure which shows the position of the inner side and outer side pixel with respect to an attention pixel. It is a figure which shows the position of the inner side and outer side pixel with respect to an attention pixel. It is a flowchart of the threshold value calculation process which concerns on embodiment. It is a flowchart of the person determination process which concerns on embodiment. It is a figure which shows the determination process of the person using the conventional person area extraction method. It is a figure which shows the determination process of the person using the person area extraction method which concerns on embodiment. It is a block diagram of information processing apparatus (computer).

Hereinafter, embodiments will be described with reference to the drawings.
FIG. 1 is a diagram illustrating a part of an image when color mixing occurs. In FIG. 1, the pixel on the left side (from the left end to the second column) is the background pixel, and the right side (from the right end to the third column) is the person area pixel. When the color mixture occurs, as shown in FIG. 1, the color of the background pixel and the color of the pixel in the person region (excluding the contour pixel) are mixed in the pixel in the human region (contour pixel) adjacent to the background pixel. Become a color.

  If the person region is extracted based on the background difference when the contour pixels of the person region are mixed colors, the mixed color pixels are extracted as the person region because they are different from the background pixels. Since the color of the mixed color pixel differs depending on the color around the person, if the mixed color pixel is included in the person region, the person characteristics obtained by the color around the person are different.

  When a surveillance camera is used in a city or a commercial facility, for example, the video is reduced in resolution so that the face cannot be grasped in order to protect the privacy of the subject. In a low-resolution image, color mixing of outline pixels of a person is likely to occur.

  In low-resolution images, the number of pixels in the person area is small, so the ratio of mixed-color pixels in the person area is large, and even in the same person image, the characteristics obtained from the person area are greatly different depending on the surrounding colors. Misjudgment of whether or not there is likely to occur.

FIG. 2 is a configuration diagram of the person determination system according to the embodiment.
The person determination system 101 includes a camera 201-i (i = 1 to 2) and a server 301.

  The cameras 201-i are installed in different places, and regularly or constantly take pictures. The camera 201-i transmits the captured image to the server 301. The camera 201-i is connected to the server 301 via a network such as a local area network (LAN) or a wide area network (WAN). In addition, the camera 201-i may be directly connected to the server 301. The number of cameras 201-i is an example and may be an arbitrary number. The camera 201-i is an example of an imaging device.

The server 301 includes a processing unit 311 and a storage unit 321.
The processing unit 311 performs processing such as extraction of a person region and determination of the same person. The processing unit 311 includes a transmission / reception unit 312, a person region extraction unit 313, a person region update unit 314, a feature generation unit 315, and a determination unit 316.

  The transmission / reception unit 312 transmits / receives data to / from a device connected to the server 301. The transmission / reception unit 312 receives an image from the camera 201-i and stores it in the storage unit 322.

  The person area extraction unit 313 detects an area (person area) where a person is shown in the image, and extracts the person area.

  When color mixing has occurred in the contour pixel, the person area update unit 314 deletes the contour pixel in which color mixing has occurred from the person area, and sets it as a new person area.

  The feature generation unit 315 calculates the feature amount of the person area and writes it in the feature information table 323. The person area extraction unit 313, the person area update unit 314, and the feature generation unit 315 are examples of a calculation unit.

The determination unit 316 determines whether or not they are the same person based on the feature amount of the person area.
The storage unit 321 stores a captured image 322-j (j = 1 to m: m is an integer of 2 or more) and a feature information table 323.

  The captured image 322-j is an image captured by the camera 201-i. Note that the transmission / reception unit 312 may receive a moving image from the camera 201-i and store it in the storage unit 322. In that case, each frame of the received moving image corresponds to the captured image 322-j.

  The feature information table 323 is a table in which feature amounts of person areas in the captured image 322-j are described. Note that the feature amount described in the feature information table 323 is the feature amount of the person area excluding the contour pixel when color mixing occurs in the person area of the captured image 322-j.

FIG. 3 is an example of a feature information table.
The feature information table 323 includes, as items, a camera ID, a person ID, coordinates 1, coordinates 2, and feature amounts. Each record of the feature information table 323 describes information about the person area of the captured image 322-j. Specifically, each record of the feature information table 323 describes a camera ID, a person ID, coordinates 1, coordinates 2, and a feature amount in association with each other. For example, in FIG. 3, the first line of the feature information table 323 describes information related to the person area of the captured image 322-1 captured by the camera 201-1, and the second line is captured by the camera 201-2. Information about the person area of the image 322-2 is described.

  The camera ID is an identifier for identifying the camera 201-i that has captured the captured image 322-j.

The person ID is an identifier for identifying a person in a person area in the captured image 322-j.
Coordinate 1 is the upper left coordinate of the rectangle that includes the person area in the captured image 322-j.

  Coordinate 2 is the lower right coordinate of the rectangle that includes the person area in the captured image 322-j. Coordinates 1 and 2 represent the approximate position of a person in the captured image 322-j.

  The feature amount is information indicating the feature of the person area. The feature amount is, for example, a color histogram of pixels in the person area. The color histogram represents the frequency of each pixel value of each color component (R (red), G (green), B (blue)) of the pixel in the person area. The color histogram may be counted for each RGB color, or the corresponding pixel may be counted for each RGB combination. Since the color histogram is compared regardless of the size (number of pixels) of the person area, the frequency is divided by the number of pixels of the person area and normalized, and the normalized frequency (normalized frequency) is stored in the feature information table 323. It is described in the feature amount. In addition, a histogram may be created by converting into a color space other than the RGB color space (for example, HSV color space), and may be used as a feature amount.

FIG. 4 is a flowchart of person area extraction processing according to the embodiment.
Here, a case will be described in which a feature amount is calculated by extracting a person region of an image (captured image 322-1) captured by the camera 201-1.

  In step S <b> 501, the transmission / reception unit 312 receives an image (captured image 322-1) captured from the camera 201-1 and stores it in the storage unit 322.

  In step S502, the person area extraction unit 313 detects a person area where a person is shown in the photographed image 322-1 and extracts the detected person area from the photographed image 322-1. The person area extraction unit 313 detects a person area by, for example, a background difference method. The person area extraction unit 313 may detect the person area using, for example, the methods described in Non-Patent Documents 1 and 2. When the background subtraction method is used, an image in which no person is shown is captured in advance by the camera 201-1, and the person area extraction unit 313 compares the image with the captured image 322-1 to be detected, and determines the difference. Based on this, the person area is detected.

  In step S503, the person area extraction unit 313 selects one of the unevaluated contour pixels as the attention pixel in the person area, and calculates a color mixture ratio of the attention pixel.

Here, a method of calculating the color mixture rate will be described.
The pixel value of a mixed color pixel is considered to be represented by the linear sum of the pixel value of the outer pixel and the pixel value of the inner pixel, and the ratio (color mixing ratio) of the pixel value of the outer pixel is equal to or greater than the threshold In addition, it is determined that the color of the person region and the background region is mixed in the contour pixel.

  FIG. 5 is a diagram illustrating a target pixel, an inner pixel, and an outer pixel. FIG. 5 is an image of a part of the person area and background of the photographed image 322-1.

  In FIG. 5, the pixels from the left end to the second column are background regions, and the pixels from the right end to the third column are human regions detected in step S502. In FIG. 5, the pixels in the person region adjacent to the background region, that is, the pixels in the third column from the left end are the contour pixels.

Here, the contour pixels of the second row and the pixel of interest I _C from the top of FIG. 5, the case of calculating the color mixing ratio of the pixel of interest I _C.

A pixel adjacent to the target pixel I _C and included in the person area is referred to as an inner pixel I _i , and a pixel adjacent to the target pixel I _C and not included in the person area is referred to as an outer pixel I _o . The pixel value of the target pixel I _C is V _c , the pixel value of the inner pixel I _i is V _i , and the pixel value of the outer pixel I _o is V _o .

Since the pixel value of the mixed color pixel is considered to be represented by the linear sum of the pixel value of the outer pixel and the pixel value of the inner pixel, the pixel value V _c of the target pixel I _C is the color mixing rate α Then, it is represented by the following formula (1).
_{V c = αV o + (1} -α) V i ··· (1)

From the above equation (1), the color mixture rate α is calculated by the following equation (2).
α = (V _c −V _i ) / (V _o −V _i ) (2)

When the pixel values V _c , V _i , and V _o are represented by three components of RGB, the pixel values V _c = (V _cR , V _cG , V _cB ), V _i = (V _iR , V _iG , When V _iB ) and V _o = (V _oR , V _oG , V _oB ), the difference between the pixel values in the above equation (2) is calculated by the Euclidean distance as follows.
V _c −V _i = ((V _cR −V _iR ) ² + (V _cG −V _iG ) ² + (V _cB −V _iB ) ² ) ^1/2
V _o −V _i = ((V _oR −V _iR ) ² + (V _oG −V _iG ) ² + (V _oB −V _iB ) ² ) ^1/2

The positions of the inner pixel and the outer pixel with respect to the target pixel are as follows, for example.
6A to 6F are diagrams illustrating the positions of the inner and outer pixels with respect to the target pixel.

As shown in FIG. 6A, if the contour pixels above and below the target pixel I _c are adjacent the target pixel among the pixels adjacent to the left and right I _c, one of the pixel next to the inner pixels I _i included in the human region The other pixel not included in the person area is the outer pixel I _o .

As shown in FIG. 6B, if the contour pixels on the left and right of the pixel of interest I _c are adjacent the target pixel among the pixels adjacent above and below the I _c, one of the pixel next to the inner pixels I _i included in the human region The other pixel not included in the person area is the outer pixel I _o .

As shown in FIG. 6C, if the contour pixels on the bottom and right of the pixel of interest I _c is adjacent, of the pixels adjacent to the lower right and upper left of the pixel of interest I _c, it is one of the pixels contained in the human region inner pixels I _i, and the other pixels that are not included in the person area is outside the pixel I _o.

As shown in FIG. 6D, if the contour pixels on the top and left of the pixel of interest I _c is adjacent, of the pixels adjacent to the upper left and lower right of the pixel of interest I _c, it is one of the pixels contained in the human region inner pixels I _i, and the other pixels that are not included in the person area is outside the pixel I _o.

As shown in FIG. 6E, if the contour pixels on the bottom and left of the pixel of interest I _c is adjacent, of the pixels adjacent to the lower left and upper right of the pixel of interest I _c, one of pixels included in the person area inside pixels I _i, and the other pixels that are not included in the person area is outside the pixel I _o.

As shown in FIG. 6F, if the contour pixels on the top and right of the pixel of interest I _c is adjacent, of the pixels adjacent to the upper right and lower left of the pixel of interest I _c, one of pixels included in the person area inside pixels I _i, and the other pixels that are not included in the person area is outside the pixel I _o.

  As the threshold used for determining whether or not colors are mixed (step S504), a preset value may be used, or may be calculated based on the background of the captured image 322-j, as will be described later.

Returning to FIG.
In step S504, the person region extraction unit 313 compares the color mixture ratio of the target pixel with a threshold value, and determines whether or not the color mixture ratio of the target pixel is equal to or greater than the threshold value. If the color mixture ratio of the target pixel is equal to or greater than the threshold, the control proceeds to step S505. If the color mixture ratio of the target pixel is less than the threshold, the control proceeds to step S506. Note that the threshold is, for example, a preset constant. Further, a threshold value calculated by a threshold value calculation process described later may be used.

  In step S505, the person area update unit 314 deletes the target pixel from the person area, and sets the person area from which the target pixel is deleted as a new person area.

  In step S506, the person area update unit 314 determines whether or not all the contour pixels have been evaluated, that is, whether or not the color mixing ratio of all the contour pixels is compared with the threshold value. If all the contour pixels have been evaluated, the control proceeds to step S507. If all the contour pixels have not been evaluated, the control returns to step S503.

  In step S507, the feature generation unit 315 calculates a feature amount of the person area. For example, the feature generation unit 315 calculates a color histogram of the person area. That is, the frequency of each pixel value of each color component (R (red), G (green), B (blue)) of the pixel in the person area is calculated. The color histogram may be counted for each RGB color, or the corresponding pixel may be counted for each RGB combination. Since color histograms are compared regardless of the size (number of pixels) of the person area, the frequency may be normalized by dividing by the number of pixels of the person area, and the normalized frequency (normalized frequency) may be used as the feature amount. . In addition, a histogram may be created by converting into a color space other than the RGB color space (for example, HSV color space), and may be used as a feature amount.

  In step S508, the feature generation unit 315 writes the calculated color histogram in the feature amount column of the feature information table 323. In addition, the feature generation unit 315 describes the camera ID, person ID, coordinates 1 and coordinates 2 in the feature information table 323. Specifically, for example, the transmission / reception unit 312 notifies the feature generation unit 315 of the camera 201-1 that has captured the captured image 322-1. The feature generation unit 315 assigns the identifier “1” corresponding to the camera 201-1. It is described in the feature information table 323 as a camera ID. The feature generation unit 315 describes an identifier different from the identifier already described as the person ID of the feature information table 323 in the feature information table 323 as a camera ID. The person area extraction unit 313 notifies the feature generation unit 315 of the upper left and lower right coordinates of the rectangle including the person area, and the feature generation unit 315 sets the upper left and lower right coordinates of the notified rectangle as coordinates 1 respectively. And the coordinates 2 are described in the feature information table 323.

In the above-described embodiment, the case where the feature amount of the captured image 322-1 is calculated has been described, but the feature amount of the captured images 322-2 to 323-j is also calculated in the same manner. Further, in the above embodiment, the color mixture ratio is calculated for the contour pixel, and deletion is performed from the person area based on the determination result of whether or not color mixing is performed.
For the pixels in the person area other than the contour pixels, the color mixture ratio may be calculated and deleted from the person area based on the determination result of whether or not the colors are mixed.

  The threshold value used for determining whether or not colors are mixed (step S504) may be a preset value, or calculated based on images taken by different cameras 201-i as described below. You may do it.

FIG. 7 is a flowchart of threshold value calculation processing according to the embodiment.
In step S511, the person region extraction unit 313 acquires the first image and the second image, and the background image (first background image) of the first image and the background image (second image) of the second image. Background image). The background image is an image in which a moving object such as a person is not reflected, and is, for example, an image obtained by deleting a person area from the captured image 322-j. When a moving object such as a person is not reflected in the first image and the second image, the first image and the second image may be used as the first background image and the second background image, respectively. The first image and the second image are images taken by different cameras 201-i. The first background image and the second background image are, for example, captured images 322-j in which a moving object such as a person is not reflected. The first and second background images are images obtained by deleting the person area from the captured image 322-j. For example, the person area extraction unit 313 acquires the captured image 322-1 captured by the camera 201-1, detects and deletes the person area of the captured image 322-1, and generates a first background image. . Furthermore, the person area extraction unit 313 acquires the captured image 322-2 captured by the camera 201-2, detects and deletes the person area of the captured image 322-2, and generates a second background image. .

  In step S512, the person region extraction unit 313 calculates a color histogram of the first background image and the second background image. The method for calculating the color histogram is the same as in step S507 described above.

  In step S513, the person region extraction unit 313 calculates the similarity between the color histograms of the first background image and the second background image. The similarity is, for example, a cross-correlation between color histograms or a histogram intersection. In addition, the person area extraction unit 313 divides the first background image and the second background image into a plurality of blocks, calculates a color histogram and a similarity for each block, and calculates a plurality of background images. The minimum value or the average value of the similarities of the blocks may be used.

In step S514, the person region extraction unit 313 calculates a threshold value based on the calculated similarity, and sets the calculated threshold value as a threshold value used in step S504. When the similarity is S, the threshold T is calculated by the following equation (3).
T = r × S (3)
Here, r is an arbitrary coefficient that takes a value between 0 and 1, and is set in advance.

  For example, the person area extraction unit 313 executes a threshold calculation process before the person area extraction process. In addition, the person area extraction unit 313 may periodically execute a threshold calculation process to change the threshold.

  When the backgrounds of the two photographed images 322-j are similar in color, there is a high possibility that the color characteristics of the persons are similar even if color mixing occurs in the person area. On the other hand, when the background is a different color, if color mixing occurs in the person area, there is a high possibility that the color characteristics of the person area are different even if the color mixing ratio is low. Therefore, it is possible to set a threshold value suitable for the shooting environment by calculating a threshold value for determining whether or not to exclude the mixed color pixel from the person region based on the similarity of the background.

FIG. 8 is a flowchart of person determination processing according to the embodiment.
Here, a case will be described in which it is determined whether or not a person shown in an image taken by a different camera 201-i is the same person. Specifically, it is determined whether or not the person shown in the shot image 322-1 shot by the camera 201-1 and the person shown in the shot image 322-2 shot by the camera 201-2 are the same person. The case where it does is demonstrated. The user or the like is instructed as to which photographed image 322-j the person to be shown in is determined.

  First, the determination unit 316 receives an instruction to determine whether the first person shown in the captured image 322-1 and the second person reflected in the captured image 322-2 are the same person.

  In step S521, the determination unit 316 selects the first person. Specifically, the determination unit 316 refers to the feature information storage table 323 and acquires the feature amount of the person area of the captured image 322-1. For example, the determination unit 316 acquires the feature amount “XXX” in the first row of the feature information storage table 323 illustrated in FIG. 3.

  In step S522, the determination unit 316 selects the second person. Specifically, the determination unit 316 refers to the feature information storage table 323 and acquires the feature amount of the person area of the captured image 322-2. For example, the determination unit 316 acquires the feature amount “YYY” in the second row of the feature information storage table 323 illustrated in FIG. 3.

  In step S523, the determination unit 316 calculates the similarity between the feature amounts of the first person and the second person. That is, the determination unit 316 calculates the similarity between the feature amount “XXX” of the person area of the captured image 322-1 and the feature amount “YYY” of the person area of the captured image 322-2. As described above, the feature amount is, for example, a color histogram. The similarity is, for example, a cross-correlation between two color histograms or a histogram intersection.

  In step S524, the determination unit 316 determines whether the calculated similarity is greater than or equal to a threshold value. If the similarity is greater than or equal to the threshold, control proceeds to step S525, and if the similarity is less than the threshold, control proceeds to step S526. Note that the threshold is, for example, a preset constant.

  In step S525, the determination unit 316 determines that the first person and the second person are the same person. That is, the determination unit 316 determines that the person shown in the shot image 322-1 and the person shown in the shot image 322-2 are the same person. The determination unit 316 stores information indicating that the person shown in the photographed image 322-1 and the person shown in the photographed image 322-2 are the same person in the storage unit 321. Further, the determination unit 316 may set the two person IDs corresponding to the two feature amounts used when it is determined that they are the same person in the feature information storage table 323 to the same value.

  In step S526, the determination unit 316 determines that the first person and the second person are not the same person. That is, the determination unit 316 determines that the person shown in the shot image 322-1 and the person shown in the shot image 322-2 are not the same person.

  In the above description, the processing is performed on the combination of the captured image 322-1 and the captured image 322-2. However, the determination unit 316 performs the above-described person determination process on all combinations of the captured image 322-j. You can go.

  In the person determination process described above, it is determined whether or not two images taken by different cameras are the same person, but whether or not the two images taken by the same camera are the same person. You may judge.

  Here, a person determination process using each of the conventional person area extraction method and the person area extraction method according to the embodiment will be described.

FIG. 9 is a diagram showing a person determination process using a conventional person region extraction method.
It is assumed that the persons shown in the captured images 322-1 and 322-2 are the same person. In addition, it is assumed that the background colors of the captured images 322-1 and 322-2 are different.

  For example, a person area extraction method using a conventional background subtraction method or the like is applied to the captured images 322-1 and 322-2 to extract a person area. At this time, contour pixels adjacent to the background are also included in the person area. Further, it is assumed that color mixing occurs in the contour pixel.

  When color mixing occurs, the color (pixel value) of the contour pixel differs depending on the background color. That is, in each of the captured images 322-1 and 322-2, even if the color of the person region (excluding the contour pixel) is the same, if the background color is different, the color of the contour pixel is also different.

  In this case, when the feature amount (color histogram) of the person region including the contour pixel is calculated, the feature amounts of the person regions of the captured images 322-1 and 322-2 are different. In particular, when the captured images 322-1 and 322-2 have a low resolution, the influence of the contour pixels is large, and thus the feature amounts of the person regions of the captured images 322-1 and 322-2 are greatly different.

  Therefore, it is determined that the person shown in the captured image 322-1 and the person shown in 322-2 are not the same person, resulting in a false determination.

  FIG. 10 is a diagram illustrating a person determination process using the person region extraction method according to the embodiment.

  It is assumed that the persons shown in the captured images 322-1 and 322-2 are the same person. In addition, it is assumed that the background colors of the captured images 322-1 and 322-2 are different.

  For example, a person area extraction method using a conventional background subtraction method or the like is applied to the captured images 322-1 and 322-2 to extract a person area. At this time, contour pixels adjacent to the background are also included in the person area. Further, it is assumed that color mixing occurs in the contour pixel.

  In the embodiment, when the contour pixel is mixed in color, the contour pixel is deleted from the person area and the person area is updated. Then, the feature amount (color histogram) of the updated human region, that is, the human region that does not include the mixed color contour pixel is calculated.

  In this case, when the feature amount (color histogram) of the person region not including the contour pixel is calculated, the feature amounts of the person region not including the contour pixel in the captured images 322-1 and 322-2 match (similar).

  For this reason, the person shown in the captured image 322-1 and the person shown in 322-2 are determined to be the same person, and can be determined correctly.

  When a color mixture occurs, the person determination system according to the embodiment calculates the feature amount of the person area excluding the mixed color pixel, and determines whether or not they are the same person based on the calculated feature amount. . Thereby, it is possible to reduce erroneous determination as to whether or not they are the same person. In particular, when the captured image has a low resolution, that is, when color mixing is likely to occur, the effect of reducing erroneous determination as to whether or not they are the same person is increased.

  In the above-described embodiment, a person is determined to be the same, but the target of the same determination is not limited to a person, and may be an arbitrary object such as a car or a motorcycle. A person, a car, a motorcycle, or the like in which a captured image is shown is an example of a subject.

FIG. 11 is a configuration diagram of an information processing apparatus (computer).
The server 301 of the embodiment can be realized by an information processing apparatus 1 as shown in FIG. 11, for example.

  The information processing apparatus 1 includes a CPU 2, a memory 3, an input unit 4, an output unit 5, a storage unit 6, a recording medium drive unit 7, and a network connection unit 8, which are connected to each other by a bus 9.

  The CPU 2 is a central processing unit that controls the entire information processing apparatus 1. The CPU 2 corresponds to the processing unit 311.

  The memory 3 is a Read Only Memory (ROM) or Random Access Memory (RAM) that temporarily stores a program or data stored in the storage unit 6 (or the portable recording medium 10) during program execution. It is memory. The CPU 2 executes the various processes described above by executing programs using the memory 3.

  In this case, the program code itself read from the portable recording medium 10 or the like realizes the functions of the embodiment.

  The input unit 4 is, for example, a keyboard, a mouse, a touch panel, a camera, or a sensor, and is used for inputting an instruction or information from a user or an operator, acquiring data used in the information processing apparatus 1, or the like.

  The output unit 5 is, for example, a display or a printer. The output unit 5 is a device that outputs an inquiry to a user or an operator and a processing result, or operates under the control of the CPU 2.

  The storage unit 6 is, for example, a magnetic disk device, an optical disk device, a tape device, or the like. The information processing apparatus 1 stores the above-described program and data in the storage unit 6 and reads them into the memory 3 and uses them as necessary. The memory 3 and the storage unit 6 correspond to the storage unit 321.

  The recording medium driving unit 7 drives the portable recording medium 10 and accesses the recorded contents. As the portable recording medium, any computer-readable recording medium such as a memory card, a flexible disk, a compact disk read only memory (CD-ROM), an optical disk, a magneto-optical disk, or the like is used. The user stores the above-described program and data in the portable recording medium 10 and reads them into the memory 3 and uses them as necessary.

  The network connection unit 8 is connected to an arbitrary communication network such as a local area network (LAN) or a wide area network (WAN), and performs data conversion accompanying communication. The network connection unit 8 transmits data to a device connected via a communication network or receives data from a device connected via a communication network.

Regarding the above embodiment, the following additional notes are disclosed.
(Appendix 1)
Determining whether the first pixel of the first region in the first image is mixed with the first background pixel adjacent to the first region;
If it is determined that the first pixel is mixed with the first background pixel, the first feature amount of the first region excluding the first pixel is calculated;
Determining whether the second pixel of the second region in the second image is mixed with the second background pixel adjacent to the second region;
When it is determined that the second pixel is mixed with the second background pixel, a second feature amount of the second region excluding the second pixel is calculated,
Causing the computer to execute a process of determining whether the subject in the first area and the subject in the second area are the same based on the first feature quantity and the second feature quantity The same judgment program characterized by
(Appendix 2)
The computer
Based on the pixel value of the first pixel, the pixel value of the first background pixel, and the pixel value of the pixel in the first region adjacent to the first pixel, a first color mixing ratio And the first pixel in the first region is mixed with the first background pixel adjacent to the first region based on whether or not the first color mixing ratio is equal to or greater than a threshold value. Whether or not
Based on the pixel value of the second pixel, the pixel value of the second background pixel, and the pixel value of the pixel in the second region adjacent to the second pixel, a second color mixing ratio And the second pixel in the second region is mixed with the second background pixel adjacent to the second region based on whether the second color mixing ratio is equal to or greater than the threshold. The same determination program as set forth in appendix 1, characterized by determining whether or not
(Appendix 3)
The first image is an image photographed by a first photographing device, and the second image is an image photographed by a second photographing device different from the first photographing device;
The computer uses the threshold based on a first background image other than the first region of the first image and a second background image of the second image other than the second region. The same determination program as described in appendix 2, characterized by:
(Appendix 4)
In the first image, it is determined whether the first pixel in the first area is mixed with the first background pixel adjacent to the first area, and the first pixel is mixed with the first background pixel. When the determination is made, the first feature amount of the first region excluding the first pixel is calculated, and the second pixel of the second region is adjacent to the second region in the second image. If the second pixel is mixed with the second background pixel, and if it is determined that the second pixel is mixed with the second background pixel, the second region of the second region excluding the second contour pixel is determined. A calculation unit for calculating a feature amount;
A determination unit configured to determine whether the subject in the first area and the subject in the second area are the same based on the first feature amount and the second feature amount;
The same determination apparatus provided with.
(Appendix 5)
The calculation unit is based on the pixel value of the first pixel, the pixel value of the first background pixel, and the pixel value of the pixel in the first region adjacent to the first pixel. A first background in which a first pixel in the first area is adjacent to the first area based on whether the first color mixing ratio is calculated and the first color mixing ratio is equal to or greater than a threshold value. A pixel value of the second pixel, a pixel value of the second background pixel, and a pixel of the pixel in the second region adjacent to the second pixel are determined. A second color mixing ratio is calculated based on the value, and a second pixel in the second region is calculated based on whether the second color mixing ratio is equal to or greater than the threshold. The identical determination device according to appendix 4, wherein it is determined whether or not a color is mixed with a second background pixel adjacent to the region.
(Appendix 6)
The first image is an image photographed by a first photographing device, and the second image is an image photographed by a second photographing device different from the first photographing device;
The calculation unit, based on a first background image other than the first region of the first image and a second background image other than the second region of the second image, The identical determination device according to appendix 5, wherein a threshold value is calculated.
(Appendix 7)
The same determination method executed by the information processing apparatus,
Determining whether the first pixel of the first region in the first image is mixed with the first background pixel adjacent to the first region;
If it is determined that the first pixel is mixed with the first background pixel, the first feature amount of the first region excluding the first pixel is calculated;
Determining whether the second pixel of the second region in the second image is mixed with the second background pixel adjacent to the second region;
When it is determined that the second pixel is mixed with the second background pixel, a second feature amount of the second region excluding the second pixel is calculated,
Based on the first feature value and the second feature value, it is determined whether or not the subject in the first area and the subject in the second area are the same. Method.
(Appendix 8)
The information processing apparatus includes:
Based on the pixel value of the first pixel, the pixel value of the first background pixel, and the pixel value of the pixel in the first region adjacent to the first pixel, a first color mixing ratio And the first pixel in the first region is mixed with the first background pixel adjacent to the first region based on whether or not the first color mixing ratio is equal to or greater than a threshold value. Whether or not
Based on the pixel value of the second pixel, the pixel value of the second background pixel, and the pixel value of the pixel in the second region adjacent to the second pixel, a second color mixing ratio And the second pixel in the second region is mixed with the second background pixel adjacent to the second region based on whether the second color mixing ratio is equal to or greater than the threshold. The same determination method according to appendix 7, characterized in that it is determined whether or not there is.
(Appendix 9)
The first image is an image photographed by a first photographing device, and the second image is an image photographed by a second photographing device different from the first photographing device;
The information processing device, based on a first background image other than the first region of the first image and a second background image other than the second region of the second image, The same determination method according to appendix 8, wherein the threshold value is calculated.

DESCRIPTION OF SYMBOLS 101 Person determination system 201 Camera 301 Server 311 Processing part 312 Transmission / reception part 313 Person area extraction part 314 Person area update part 315 Feature generation part 316 Determination part 321 Storage part 322 Captured image 323 Feature information storage table

Claims (5)

Determining whether the first pixel of the first region in the first image is mixed with the first background pixel adjacent to the first region;
If it is determined that the first pixel is mixed with the first background pixel, the first feature amount of the first region excluding the first pixel is calculated;
Determining whether the second pixel of the second region in the second image is mixed with the second background pixel adjacent to the second region;
When it is determined that the second pixel is mixed with the second background pixel, a second feature amount of the second region excluding the second pixel is calculated,
Causing the computer to execute a process of determining whether the subject in the first area and the subject in the second area are the same based on the first feature quantity and the second feature quantity The same judgment program characterized by The computer
Based on the pixel value of the first pixel, the pixel value of the first background pixel, and the pixel value of the pixel in the first region adjacent to the first pixel, a first color mixing ratio And the first pixel in the first region is mixed with the first background pixel adjacent to the first region based on whether or not the first color mixing ratio is equal to or greater than a threshold value. Whether or not
Based on the pixel value of the second pixel, the pixel value of the second background pixel, and the pixel value of the pixel in the second region adjacent to the second pixel, a second color mixing ratio And the second pixel in the second region is mixed with the second background pixel adjacent to the second region based on whether the second color mixing ratio is equal to or greater than the threshold. The same determination program according to claim 1, wherein it is determined whether or not there is any. The first image is an image photographed by a first photographing device, and the second image is an image photographed by a second photographing device different from the first photographing device;
The computer uses the threshold based on a first background image other than the first region of the first image and a second background image of the second image other than the second region. The same determination program according to claim 2, wherein: In the first image, it is determined whether the first pixel in the first area is mixed with the first background pixel adjacent to the first area, and the first pixel is mixed with the first background pixel. If it is determined, the first feature amount of the first region excluding the first pixel is calculated, and the second pixel of the second region in the second image is the second region and the second region. If it is determined whether the second background pixel is mixed with a color, and if it is determined that the second pixel is mixed with the second background pixel, the second region of the second region excluding the second pixel A calculation unit for calculating a feature amount;
A determination unit configured to determine whether the subject in the first area and the subject in the second area are the same based on the first feature amount and the second feature amount;
The same determination apparatus provided with. The same determination method executed by the information processing apparatus,
Determining whether the first pixel of the first region in the first image is mixed with the first background pixel adjacent to the first region;
If it is determined that the first pixel is mixed with the first background pixel, the first feature amount of the first region excluding the first pixel is calculated;
Determining whether the second pixel of the second region in the second image is mixed with the second background pixel adjacent to the second region;
When it is determined that the second pixel is mixed with the second background pixel, a second feature amount of the second region excluding the second pixel is calculated,
Based on the first feature value and the second feature value, it is determined whether or not the subject in the first area and the subject in the second area are the same. Method.