JP2022175957A - Person authentication device, person authentication method, person authentication program and person authentication system - Google Patents

Abstract

To suppress a failure of authentication of a person in a video that is shot at night.SOLUTION: A person authentication device comprises: an acquisition unit which acquires a type of a light source corresponding to a position where image data is captured; a collection unit which collects a correspondence between a color of a possessed article of an object person in first image data captured under a first type light source and a color of a possessed article of the object person in second image data captured under a second type light source; and an authentication unit which calculates a hue possibility indicating such a possibility that a comparison person whose possessed article is in the second color in the second image data is the object person when the possessed article of the object person in the first image data is in the first color on the basis of the correspondence.SELECTED DRAWING: Figure 8

Description

The present invention relates to a person authentication device, a person authentication method, a person authentication program, and a person authentication system.

2. Description of the Related Art In recent years, in applications such as crime prevention, there is an increasing demand for person authentication technology for authenticating a specific person from images captured by security cameras and the like. Systems using human authentication technology include, for example, a person matching system that analyzes images captured by cameras installed around roads, detects the target person, and tracks the target person on a continuous image. . For example, the police take a suspect of a certain crime as a target person, and track the target person by associating the same person from images of security cameras installed at a plurality of locations. For authentication of the target person, for example, face authentication that identifies the target person from facial features is used.

Techniques for identifying a person from an image are disclosed below.

JP-A-2005-332071 Japanese Patent Application Laid-Open No. 2010-147560

However, for example, in an image taken at night when crime is common, even the same person appears differently depending on the location where the image was taken due to surrounding lights such as security lights. In this case, detection of the same person may fail.

Accordingly, one disclosure provides a person authentication device, a person authentication method, a person authentication program, and a person authentication system that suppress failures in person authentication in images shot at night.

an acquisition unit that acquires the type of light source corresponding to the position where the image data was captured; the color of the belongings of the target person in the first image data captured under the first type light source; a collection unit that collects a correspondence relationship between the color of the subject's belongings in the photographed second image data and the color of the subject's belongings in the first image data; an authentication unit that calculates a color probability indicating a probability that the target person is a comparison person whose belongings in the two-image data are the second color based on the correspondence relationship.

One disclosure can suppress failures in person authentication in video shot at night.

FIG. 1 is a diagram showing a configuration example of a person authentication system 10. As shown in FIG. FIG. 2 is a diagram showing a configuration example of the image processing apparatus 100. As shown in FIG. FIG. 3 is a diagram showing an example of a processing flowchart of the person tracking processing S100. FIG. 4 is a diagram showing an example of a processing flowchart of the color information collection processing S200. FIG. 5 is a diagram showing an example of a locus of movement of a target person. FIG. 6 is a diagram showing an example of an image of a target person at each point. FIG. 7 is a diagram showing an example of a security light map. FIG. 8 is a diagram showing an example of a processing flowchart of the image-by-image target person authentication processing S300. FIG. 9 is a diagram showing an example of color data. FIG. 10 is a diagram showing an example of a processing flowchart of the image-by-image target person authentication processing S400 according to the second embodiment. FIG. 11 is a diagram showing an example of color data that stores correct colors. FIG. 12 is a diagram showing an example of color data corresponding to multiple colors.

[First embodiment]
A first embodiment will be described.

<Configuration example of person authentication system 10>
FIG. 1 is a diagram showing a configuration example of a person authentication system 10. As shown in FIG. The person authentication system 10 has an image processing device (person authentication device) 100, cameras 200-1 to 200-3, and a network 300. FIG. The person authentication system 10 is a system that analyzes image data of videos taken by the cameras 200-1 to 200-3 and searches for a specific person.

Cameras 200-1 to 200-3 (hereinafter sometimes referred to as cameras 200) are photographing devices for photographing people and the like. The camera 200 is, for example, a monitoring camera installed on roads, houses, or the like. The camera 200 is connected to the image processing apparatus 100 via the network 300 and transmits image data of the captured video to the image processing apparatus 100 . For example, the camera 200 captures a moving image as an image, and continuously transmits moving image data of a predetermined size or predetermined time to the image processing apparatus 100 . In addition, the camera 200 shoots moving images and still images all the time or periodically.

Target persons 400-1 to 400-3 (hereinafter sometimes referred to as target persons 400) are persons to be searched. The target person 400 moves in the moving direction D1 over time. Then, the target person 400 is photographed by the camera 200 at the destination.

The image processing apparatus 100 is, for example, a computer or a server machine that analyzes image data acquired from the camera 200 and detects a target person 400 appearing in the image data.

Network 300 is, for example, the Internet. Also, the network may be a local network or a wireless network.

In the person authentication system 10 , a camera 200 photographs a moving target person 400 . The image processing device 100 searches for the target person 400 from the image data.

The image processing apparatus 100 detects a person from video data and authenticates whether or not the detected person is a target person. The image processing apparatus 100 performs authentication (first authentication) by comparing the detected person with the appearance of the target person (excluding the color of the clothing). Then, in addition to the first authentication, the image processing apparatus 100 performs authentication (second authentication) using the color of the target person's clothing (belongings). In the second authentication, the image processing apparatus 100 considers the color change due to the security light (light source, light) at the image capturing position. The image processing device 100 collects the color change caused by the security light from the image data acquired before or during the initial period of tracking the target person, for example. The image processing apparatus 100 improves the accuracy of person authentication by performing the second authentication.

<Configuration Example of Image Processing Apparatus 100>
FIG. 2 is a diagram showing a configuration example of the image processing apparatus 100. As shown in FIG. The image processing apparatus 100 has a CPU (Central Processing Unit) 110 , a storage 120 , a memory 130 , a communication circuit 140 and an accelerator 150 .

The storage 120 is an auxiliary storage device such as flash memory, HDD (Hard Disk Drive), or SSD (Solid State Drive) that stores programs and data. The storage 120 stores a person detection program 121 , a person tracking program 122 and a security light map 123 .

The security light map 123 is, for example, light source information such as a map or a position information list that stores the correspondence relationship between the shooting position of the image and the type of security light (light source) that illuminates the vicinity of the shooting position.

Memory 130 is an area into which programs stored in storage 120 are loaded. The memory 130 may also be used as an area where programs store data.

The CPU 110 is a processor that loads a program stored in the storage 120 into the memory 130, executes the loaded program, constructs each section, and realizes each process.

The communication circuit 140 is a circuit that communicates with the camera 200 via the network 300 . The communication circuit 140 is, for example, a NIC (Network Interface Card). Communication circuit 140 may be connected to network 300 either wirelessly or by wire.

The accelerator 150 is, for example, a GPU (Graphics Processing Unit) or FPGA (Field Programmable Gate Array) that performs specific processing according to instructions from the CPU 110 . The accelerator 150 is a hardware accelerator, and performs image processing, arithmetic processing, and the like, for example.

The CPU 110 performs person detection processing by executing the person detection program 121 . Person detection processing is processing for detecting a person from image data. The image processing apparatus 100 analyzes (image processes) the acquired image data and detects a person appearing in the image.

By executing the person tracking program 122, the CPU 110 constructs an acquisition unit, a collection unit, and an authentication unit, and performs person tracking processing. Person tracking processing is processing for detecting a target person from image data. The image processing apparatus 100 analyzes (image processes) the acquired image data and detects a target person appearing in the image. The person tracking process includes, for example, color information collection process and image-by-image target person authentication process.

The CPU 110 performs color information collection processing by executing the color information collection module 1221 of the person tracking program 122 . The color information collection process is a process of collecting color data. The color data is data that stores the correspondence relationship between the type of security light (light source) and the color of the target person's clothes (belongings). The color data is, for example, data that stores what color appears under a security light (what color changes) under another type of security light.

The CPU 110 performs image-by-image target person authentication processing by executing the image-by-image target person authentication module 1222 of the person tracking program 122 . The target person authentication process for each image calculates the probability that the person in the image is the target person based on the authentication that does not use color data (the color of the target person's clothing is not used) and the second authentication based on the color data. It is a process to Further, the image-by-image target person authentication process may be a process of determining whether or not a person appearing in an image is a target person.

<About person tracking processing>
Person tracking processing will be described. The person tracking process is, for example, a process of tracking a target person such as a criminal or a missing person using an image captured by the camera 200 .

FIG. 3 is a diagram showing an example of a processing flowchart of the person tracking processing S100. The image processing apparatus 100 identifies a target person in the person tracking process S100 (S100-1). A target person is a person to be tracked.

The image processing apparatus 100 performs color information collection processing (S200). The color information collection process S200 is a process of collecting the colors of belongings such as clothes of the target person, and creating a database in association with security light types. The color is information about the color of the belongings worn by the target person, and includes, for example, the color of clothes such as jackets, trousers, and skirts. Also, the color includes, for example, the color of hats, shoes, gloves, and the like. The color information collection processing S200 will be described later.

When the color information collection processing S200 is completed, the image processing apparatus 100 acquires the next image (S100-2), and performs the target person authentication processing for each image (S300). The image-by-image target person authentication process S300 is a process of determining whether or not a person appearing in an image is a target person (or calculating the degree or probability of being a target person) based on color information and other appearance information. be. The image-by-image target person authentication processing S300 will be described later.

After completion of the image-by-image target person authentication process S300, the image processing apparatus 100 determines whether or not the tracking end condition is satisfied (S100-3), and if the tracking end condition is satisfied (Yes in S100-3), the process is started. finish. On the other hand, if the tracking end condition is not satisfied (No in S100-3), the image processing apparatus 100 acquires the image again (S100-2), and executes the target person authentication process S300 for each image.

The tracking end condition is a condition for ending the tracking of the target person. The tracking end condition may be, for example, a condition based on the authentication result of the target person authentication process S300 for each image, such as the target person being undetectable (lost). Also, the tracking end condition may be a physical condition such as tracking time or the presence or absence of the next image. Furthermore, the tracking end condition may be a condition related to the necessity or possibility of tracking, such as disappearance of the need to track the target person or entering a building where the target person cannot be photographed.

<Color information collection processing>
FIG. 4 is a diagram showing an example of a processing flowchart of the color information collection processing S200. The image processing apparatus 100 acquires an image (S200-1). The image processing apparatus 100 searches for the target person from the image (S200-2). In addition, in the search in step S200-2, the person is not specified by color such as the color of the clothes. In step S200-2, the image processing apparatus 100 detects, for example, characteristic belongings such as a bag and a hat possessed by the target person, and associates persons who have the same belongings to search for the target person. . Henceforth, the data collected in color information collection process S200 may be called color data. Also, hereinafter, the object to be collected as color data will be explained by taking the color of the object person's clothes (especially the jacket) as an example, but the objects worn by the object person such as pants, shoes, hats, bags, skin, etc. It may be a color or a color of a part of the target person.

When the target person is not detected (No in S200-3), the image processing apparatus 100 acquires the next image (S200-1). On the other hand, when the target person is detected (Yes in S200-2), the image processing apparatus 100 extracts the type of security light corresponding to the shooting position of the image (S200-4). The image processing apparatus 100 identifies and extracts the type of security light from the security light map 123, for example.

The image processing apparatus 100 identifies the color of the target person's clothes (S200-5). Colors may be specified by color names such as “red”, “yellow”, and “RED”, or may be specified by numerical values such as the RGB color model, YUV system, YCbCr system, and YPbPr system. good.

The image processing apparatus 100 aggregates the information regarding the specified color as color data (S200-6).

The image processing apparatus 100 terminates the process if the termination condition is satisfied (Yes in S200-7). On the other hand, if the termination condition is not satisfied (No in S200-7), the image processing apparatus 100 acquires the next image (S200-1). The termination condition is a condition for terminating the color information collection processing S200, such as the amount (number) of color data collected, the time to execute the color information collection processing S200, the number of processed images, or the time. It is a condition according to

An example of the color information collection processing S200 will be described below using an image and a security light map.

FIG. 5 is a diagram showing an example of a locus of movement of a target person. The target person walks from point P1 to point P3 via point P2. Then, the target person is photographed, for example, by a security camera or the like at the point P1, the point P2, and the point P3.

FIG. 6 is a diagram showing an example of an image of a target person at each point. The image G1 is an image of the target person photographed at the point P1. The image G2 is an image of the target person photographed at the point P2. The image G3 is an image of the target person photographed at the point P3.

In the color information collection process S200, the image processing apparatus 100 acquires the image G1 (S200-1 in FIG. 4) and identifies the target person (Yes in S200-2 and S200-3 in FIG. 4). The image processing apparatus 100 extracts the type of security light corresponding to the shooting position of the image G1 from the security light map (S200-4).

FIG. 7 is a diagram showing an example of a security light map. FIG. 7 shows that the area A1 is the area illuminated by the security light A, and the area A2 is the area illuminated by the security light B. FIG. The image processing apparatus 100 extracts from the security light map that the spot P1 where the image G1 was taken is illuminated by the security light A (S200-4 in FIG. 4).

In the color information collection process S200, the image processing apparatus 100 extracts the color of the clothes of the target person in the image G1 (S200-5 in FIG. 4). The image processing apparatus 100 extracts the color of the target person's clothes as "red" from the image G1 in FIG. 6 (S200-5 in FIG. 4).

The image processing apparatus 100 associates the color "red" of the target person's clothes with the crime prevention light A, and aggregates them as color data (S200-6 in FIG. 4). Color data is created for each type of security light, for example. The color data indicates, for example, the relationship between the color under a security light and the color under another security light. The image processing apparatus 100 stores, as color data, the fact that the color of clothing that is "red" under the security light A appears "orange" under the security light B (or the probability that it appears). .

The image processing apparatus 100 acquires the next image G2 (S200-1 in FIG. 4) and identifies the target person (Yes in S200-2 and S200-3 in FIG. 4). The image processing apparatus 100 extracts the type of security light corresponding to the shooting position of the image G2 from the security light map (S200-4).

The image processing apparatus 100 extracts from the security light map that the point P2 where the image G2 was captured is illuminated by the security light B (S200-4 in FIG. 4).

The image processing apparatus 100 extracts the color of the target person's clothes as "orange" from the image G2 in FIG. 6 (S200-5 in FIG. 4).

The image processing apparatus 100 associates the color "orange" of the target person's clothes with the crime prevention light B, and aggregates them as color data (S200-6 in FIG. 4).

Further, the image processing apparatus 100 acquires the next image G3 (S200-1 in FIG. 4) and identifies the target person (Yes in S200-2 and S200-3 in FIG. 4). The image processing apparatus 100 extracts the type of security light corresponding to the shooting position of the image G2 from the security light map (S200-4).

The image processing apparatus 100 extracts from the security light map that the point P3 where the image G3 was taken is illuminated by the security light B (S200-4 in FIG. 4).

The image processing apparatus 100 extracts the color of the target person's clothes as "orange" from the image G3 in FIG. 6 (S200-5 in FIG. 4).

The image processing apparatus 100 associates the color "orange" of the target person's clothes with the crime prevention light B, and aggregates them as color data (S200-6 in FIG. 4).

As described above, the image processing apparatus 100 acquires images until the termination condition is satisfied, associates the color of the target person's clothes with the type of the security light, and stores them as color data.

Note that the color information collection process S200 may be executed for all images processed in the execution of the person tracking process S100, for example, without setting an end condition. For example, in the person tracking process S100, if the target person can be specified (detected) without using information based on the color of the clothes, the image processing apparatus 100 aggregates the color of the target person's clothes in the specified image into color data. You may Accordingly, as the amount of color data increases and the execution time of the person tracking process S100 increases, the accuracy of authentication of the target person based on the color data improves.

<Per-image target person authentication processing>
FIG. 8 is a diagram showing an example of a processing flowchart of the image-by-image target person authentication processing S300. The image processing apparatus 100 extracts the type of security light at the shooting position of the image (S300-1).

If an undetermined person exists in the image (Yes in S300-2), the image processing apparatus 100 extracts the person from the image (S300-3). The image processing apparatus 100 executes the process S300-3 by, for example, detecting all persons in the image and selecting any one of the detected persons (or in order of clarity of the image).

The image processing apparatus 100 calculates a first probability that the extracted person is the target person (S300-4). The image processing apparatus 100 does not make determination based on the color of the clothes in step S300-4. The first probability (person probability) is, for example, the probability of the first authentication (person authentication), and is a probability based on physical features such as the face and height of the extracted person. Also, the first probability is, for example, a probability based on what the extracted person is wearing (such as a hat) or belongings (such as a bag). The first probability is a probability calculated without using information about the color of the clothes of the extracted person.

The image processing apparatus 100 calculates a second probability (color probability) that the extracted person is the target person based on the color of the extracted person's clothing. The second probability is, for example, the probability of second authentication, and is the probability calculated (extracted) based on the color of the extracted person's clothes and color data.

The image processing apparatus 100 determines whether the extracted person is the target person based on the first probability and the second probability (S300-6). The image processing apparatus 100 may calculate the probability (comprehensive probability) that the extracted person is the target person in step S300-6. For example, the image processing apparatus 100 considers (determines) that the extracted person is the target person when the integrated probability is equal to or greater than the threshold.

When the image processing apparatus 100 determines that the extracted person is the target person (Yes in S300-7), the image processing apparatus 100 identifies the extracted person as the target person (S300-8), and ends the process.

If the image processing apparatus 100 determines that the extracted person is not the target person (No in S300-7), or if an undetermined person exists in the image (Yes in S300-2), the image processing apparatus 100 extracts the next person ( S300-3).

If the undetermined person does not exist in the image (No in S300-2), the image processing apparatus 100 determines that the target person is not in the image (S300-9), and ends the process.

Note that the image processing apparatus 100 does not need to determine whether each extracted person is a target person (process S300-7), for example, when calculating the integration probability in the process S300-6. In this case, the image processing apparatus 100 identifies the extracted person with the highest integration probability among the determined extracted persons as the target person. In addition, the image processing apparatus 100 may select an extracted person whose integrated probability is higher than a threshold as a candidate for the target person, among the determined extracted persons.

FIG. 9 is a diagram showing an example of color data. The color data in FIG. 9 is data listing the probabilities of what color the color under the security light A will appear under the security light B (what color it will change to). For example, according to FIG. 9, the probability that clothing colored "red" under security light A will appear (change) to be "orange" under security light B is 75%, which is the highest probability. Recognize. The probability in FIG. 9 is calculated from the degree of appearance of the color "red" under the security light A in the images shot under the plurality of security lights B in the color information collection process S200. value. FIG. 9 shows, for example, that out of 100 images taken under security light B, 75 were orange.

The image processing apparatus 100 calculates the second probability using the color data in FIG. 9 in the image-by-image target person authentication process S300 (S300-5 in FIG. 8). The image processing device 100 sets the second probability to 75% when the color of the clothes of the target person is red under the security light A and the color of the clothes of the extracted person in the image data of the security light B is orange. to calculate. Then, the image processing apparatus 100 uses the first probability and the second probability to calculate the probability (integrated probability) that the extracted person is the target person, and performs authentication (S300-6, S300-7 in FIG. 8). .

In the first embodiment, the image processing apparatus 100 collects the color of the target person's clothes for each security light type. Then, the image processing apparatus 100 calculates the probability that the person is the target person (second probability) based on the probability of the change in the color of the clothes (probability of the color corresponding to the change in the type of the security light), and determines that the person is the target person. Determine whether or not the person is a person, or calculate the probability. As a result, the image processing apparatus 100 can detect the target person in consideration of the change in the color of the clothes depending on the type of the security light, and can suppress erroneous recognition of the target person.

[Second embodiment]
A second embodiment will be described. In the second embodiment, the image processing apparatus 100 extracts from the image a person whose clothing color matches the target person (or the probability of matching is higher than a predetermined value) in the image-by-image target person authentication process S300, Calculate the probability that the selected person is the target person.

In the second embodiment, since the person to be authenticated by the color of the clothes is extracted in advance, it is possible to exclude other people whose clothes are clearly different in color from the authentication targets, thereby suppressing an increase in the amount of processing. be able to.

<Per-image target person authentication processing>
FIG. 10 is a diagram showing an example of a processing flowchart of the image-by-image target person authentication processing S400 according to the second embodiment. The image processing apparatus 100 extracts the type of security light at the image capturing position (S400-1).

The image processing apparatus 100 extracts a person whose clothes match the color of the person from the image (S400-2). Matching the colors of the clothes means, for example, matching the colors of the change under the security light. For example, in the case of the color data of FIG. 9, the image processing apparatus 100 extracts the person wearing orange clothes with the highest probability from the image captured under the security light B. FIG. Alternatively, the image processing apparatus 100 may extract a person wearing orange and black clothes whose probability is equal to or greater than a predetermined value (for example, 10%). The image processing apparatus 100 extracts a person with a color that has a high probability.

If an undetermined person exists in the image (Yes in S400-3), the image processing apparatus 100 selects the person from the image (S400-4).

The image processing apparatus 100 calculates the probability that the selected person is the target person (S400-4). The image processing device 100 calculates, for example, the first probability.

The image processing apparatus 100 determines whether or not the selected person is the target person based on the calculated probability (S400-6). For example, the image processing apparatus 100 considers (determines) that the selected person is the target person when the probability is equal to or greater than the threshold.

When the image processing apparatus 100 determines that the selected person is the target person (Yes in S400-7), the image processing apparatus 100 identifies the selected person as the target person (S400-8), and ends the process.

If the image processing apparatus 100 determines that the selected person is not the target person (No in S400-7), or if an undetermined person exists in the image (Yes in S400-2), it selects the next extracted person. (S400-3).

When the undetermined person does not exist in the image (No in S400-3), the image processing apparatus 100 determines that the target person is not in the image (S400-8), and ends the process.

[Other embodiments]
Color data may store the color with the highest probability of change (correct color) in addition to the probability of change. FIG. 11 is a diagram showing an example of color data that stores correct colors. According to FIG. 11, for example, the color "orange" under the security light A appears "red" under the security light B, and "black" under the security lights C and D.

Alternatively, the color data may be a combination of multiple color change probabilities. FIG. 12 is a diagram showing an example of color data corresponding to multiple colors. FIG. 12 shows an example of the probability of change in the security light B when there are multiple target persons or when the color of the target person's clothes appears in a plurality of different colors even under the same security light A. It is a diagram. By storing the color data in this manner, a plurality of target persons, for example, a person wearing orange clothes and a person wearing blue clothes, can be detected from the security light B image. When a plurality of colors are stored with probability as shown in FIG. 12, the image processing apparatus 100, in tracking the target person, selects the correct color (for example, the color of the target person's clothes under a security light), A target person may be tracked by determining a color with a high appearance probability and searching for a person based on the correct color.

The following is summarized as an appendix.

(Appendix 1)
an acquisition unit that acquires the type of light source corresponding to the position where the image data was captured;
the color of the target person's belongings in the first image data photographed under the first type light source and the color of the target person's belongings in the second image data photographed under the second type light source, a collection unit that collects correspondence relationships;
When the belongings of the target person in the first image data are the first color, the color probability indicating the probability that the comparison person whose belongings are the second color in the second image data is the target person an authentication unit that calculates based on the correspondence;
A person authentication device having

(Appendix 2)
The person authentication device according to Supplementary Note 1, wherein the correspondence relationship includes a probability that the first color of the first image data changes to the second color of the second image data.

(Appendix 3)
The person authentication device according to Supplementary Note 1, wherein the collecting unit identifies the target person by person authentication that does not use the color of the belongings, and collects the correspondence.

(Appendix 4)
The collecting unit regards belongings of a person who has been authenticated as being the target person with a probability equal to or higher than a predetermined probability in person authentication that does not use the color of the belongings, and collects the correspondence relationship. The person authentication device according to Supplementary Note 1.

(Appendix 5)
The authentication unit further performs the person authentication on the comparison person, and determines that the comparison person is the target person based on the person probability that the comparison person is the target person and the color probability in the person authentication. The person authentication device according to appendix 3, which calculates a probability that the person is a person.

(Appendix 6)
The person authentication device according to Supplementary Note 1, wherein the correspondence includes color change information indicating to what color the first color of the first image data changes to a color of the second image data.

(Appendix 7)
The person authentication device according to supplementary note 1, further comprising a data acquisition unit that acquires the image data from a photographing device.

(Appendix 8)
having light source information indicating correspondence between shooting positions and light source types;
The person authentication device according to supplementary note 1, wherein the acquisition unit acquires the type of the light source from the light source information.

(Appendix 9)
The person authentication device according to appendix 1, wherein the personal belongings include clothes of the person.

(Appendix 10)
Acquire the type of light source corresponding to the position where the image data was captured,
the color of the target person's belongings in the first image data photographed under the first type light source and the color of the target person's belongings in the second image data photographed under the second type light source, collect correspondences,
When the belongings of the target person in the first image data are the first color, the color probability indicating the probability that the comparison person whose belongings are the second color in the second image data is the target person Person authentication method calculated based on correspondence.

(Appendix 11)
A process of acquiring the type of light source corresponding to the position where the image data was captured;
the color of the target person's belongings in the first image data photographed under the first type light source and the color of the target person's belongings in the second image data photographed under the second type light source, a process of collecting correspondences;
When the belongings of the target person in the first image data are the first color, the color probability indicating the probability that the comparison person whose belongings are the second color in the second image data is the target person a process of calculating based on the correspondence;
A person authentication program that makes a computer run

(Appendix 12)
a plurality of imaging devices for capturing images;
Acquiring the image data photographed by the photographing device, acquiring the type of light source corresponding to the position where the image data was photographed, and obtaining the belongings of the target person in the first image data photographed under the first type light source. A correspondence relationship between a color and a color of the belongings of the target person in the second image data photographed under a second type light source is collected, and the belongings of the target person in the first image data are the first a person authentication device for calculating, based on the correspondence relationship, a color probability indicating a probability that a comparative person whose property in the second image data is the second color is the target person when the color is a color;
A person authentication system having

10: person authentication system 100: image processing device 110: CPU
120 : Storage 121 : Person detection program 122 : Person tracking program 1221 : Color information collection module 1222 : Image-by-image target person authentication module 123 : Security light map 130 : Memory 140 : Communication circuit 150 : Accelerator 200 : Camera 300 : Network 400 : Target person

Claims (10)

an acquisition unit that acquires the type of light source corresponding to the position where the image data was captured;
the color of the target person's belongings in the first image data photographed under the first type light source and the color of the target person's belongings in the second image data photographed under the second type light source, a collection unit that collects correspondence relationships;
When the belongings of the target person in the first image data are the first color, the color probability indicating the probability that the comparison person whose belongings are the second color in the second image data is the target person an authentication unit that calculates based on the correspondence;
A person authentication device having 2. The person authentication device according to claim 1, wherein the correspondence relationship includes a probability that the first color of the first image data changes to the second color of the second image data. 2. The person authentication device according to claim 1, wherein the collecting unit identifies the target person by person authentication that does not use the color of the belongings, and collects the correspondence. The collecting unit regards belongings of a person who has been authenticated as being the target person with a probability equal to or higher than a predetermined probability in person authentication that does not use the color of the belongings, and collects the correspondence relationship. The person authentication device according to claim 1. The authentication unit further performs the person authentication on the comparison person, and based on the person probability indicating the probability that the comparison person is the target person in the person authentication and the color probability, the comparison person is the target person. 2. The person authentication device according to claim 1, wherein the correspondence relationship includes color change information indicating to what color the first color of the first image data changes to what color in the second image data. having light source information indicating correspondence between shooting positions and light source types;
The person authentication device according to claim 1, wherein the acquisition unit acquires the type of the light source from the light source information. Acquire the type of light source corresponding to the position where the image data was captured,
the color of the target person's belongings in the first image data photographed under the first type light source and the color of the target person's belongings in the second image data photographed under the second type light source, collect correspondences,
When the belongings of the target person in the first image data are the first color, the color probability indicating the probability that the comparison person whose belongings are the second color in the second image data is the target person Person authentication method calculated based on correspondence. A process of acquiring the type of light source corresponding to the position where the image data was captured;
the color of the target person's belongings in the first image data photographed under the first type light source and the color of the target person's belongings in the second image data photographed under the second type light source, a process of collecting correspondences;
When the belongings of the target person in the first image data are the first color, the color probability indicating the probability that the comparison person whose belongings are the second color in the second image data is the target person a process of calculating based on the correspondence;
A person authentication program that makes a computer run a plurality of imaging devices for capturing images;
Acquiring the image data photographed by the photographing device, acquiring the type of light source corresponding to the position where the image data was photographed, and obtaining the belongings of the target person in the first image data photographed under the first type light source. A correspondence relationship between a color and a color of the belongings of the target person in the second image data photographed under a second type light source is collected, and the belongings of the target person in the first image data are the first a person authentication device for calculating, based on the correspondence relationship, a color probability indicating a probability that a comparative person whose property in the second image data is the second color is the target person when the color is a color;
A person authentication system having

Images