JP2023125420A - Image determination device, passage management device and image determination method - Google Patents

Abstract

To determine the presence/absence of impersonation with a minimum number of cameras.SOLUTION: An image determination device comprises: a plurality of cameras 1A, 1B which image a prescribed point of a detection object region in mutually-different visual angles; an image acquisition function unit 2 which acquires object images included in image data imaged by the cameras 1A, 1B; a correlation detection function unit 3 which detects a correlation of positions of the detection object regions of the object images acquired by the image acquisition function unit 2 from the image data imaged by the one camera 1A and the other camera 1B; and an impersonation determination function unit 4 which determines whether or not the object images in the image data of the one camera 1A and the other camera 1B are impersonation images from the object images having the correlation detected by the correlation detection function unit 3.SELECTED DRAWING: Figure 1

Description

The present invention relates to an image determination device, a traffic management device, and an image determination method.

There is Patent Document 1 as a technology related to spoofing detection in face authentication. In Patent Document 1, there are two types of images: an image obtained from a first camera, and an image obtained from a second camera that shoots from a direction that is difficult to obtain from the first camera. Analysis is underway.
Furthermore, in Patent Documents 2 and 3 related to the present invention, the feature points are mainly shared by converting feature points from images of multiple cameras into images in three-dimensional space to perform image analysis with higher accuracy. Efforts are being made to prevent identity theft by companions.

Japanese Patent Application Publication No. 2008-108243 International Publication No. 2021/060256 Japanese Patent Application Publication No. 2004-192378

However, the image analysis techniques described in Patent Documents 1 to 3 above require processing to generate a three-dimensional image from images obtained from multiple cameras, so it takes time to generate the image. It can be suitably used for active authentication such as door unlocking in building security management, but it is not practical for non-active authentication such as walk-through authentication that requires high-speed processing without interfering with the flow of people. The problem is that it is difficult to convert.
Other than Patent Documents 1 to 3 mentioned above, as a method for preventing spoofing by capturing an image of a display device or the like into a camera, a camera that is easily recognized by the target person (the target person is likely to point the fake image at the camera) and a recognition method are used. There is also a method that uses images from a camera that is paired with a camera that is difficult to detect (a hidden camera that makes it difficult for the target to point the fake image at the camera), and uses the facial recognition results from the images from both cameras to determine whether it is spoofing. However, in the case of a two-way traffic facility with two lanes of one-way walk-through gates, a total of two pairs of cameras, one pair for each lane, would be required, and furthermore, authentication processing would be required for the two pairs of camera images. As the amount of processing increases due to this, there are problems in that the processing time increases and it becomes necessary to prepare hardware with higher performance.

The purpose of this invention is to determine impersonation using a limited number of cameras.

In order to solve the above problems, the present invention proposes the following means.
The image determination device according to the first aspect of the present invention is characterized by the following configuration.
This image determination device includes a plurality of cameras that photograph predetermined points in a detection target area at different angles of view, an image detection function unit that detects target images included in image data photographed by these cameras, and a single camera. a correlation detection function unit that detects a correlation between positions in the detection target area of target images respectively detected by the image detection function unit from image data taken by the camera and another camera; and a correlation detected by the correlation detection function unit. and a spoofing determination function unit that determines whether or not a target image in image data of the one camera and another camera is a spoofed image from a target image having a spoofed image.

The image determination method according to the second aspect of the present invention is characterized by the following configuration.
This image determination method includes a step of photographing a predetermined point in the detection target region L using a plurality of cameras having different angles of view, a target image detection step of detecting a target image included in the photographed image data, and a step of detecting a target image included in the photographed image data. Detecting the correlation between the positions of the detected target images in the detection target area from the image data taken by the camera and the other camera; and determining whether the target image in the image data is a spoofed image.

According to the present invention, spoofing can be determined based on images obtained from at least two cameras.

FIG. 2 is a block diagram of an example of the minimum configuration of an image determination device according to the present invention. 1 is a flowchart of a minimum configuration example of an image determination method according to the present invention. FIG. 1 is an overall configuration diagram of an embodiment of the present invention. FIG. 2 is a block diagram of one embodiment. FIG. 2 is a block diagram at the time of preparation for determination according to an embodiment. FIG. 2 is a block diagram at the time of correlation detection according to an embodiment. FIG. 2 is a block diagram of an embodiment during standby. It is a block diagram at the time of passerby detection of one embodiment. FIG. 3 is a block diagram at the time of data association in one embodiment. FIG. 2 is a block diagram at the time of determination according to an embodiment. FIG. 3 is an explanatory diagram of rectangle association according to an embodiment. FIG. 3 is an explanatory diagram of correlation determination according to an embodiment.

The configuration of an image determination device according to the minimum configuration of the present invention will be described with reference to FIG.
This device includes a plurality of cameras 1A and 1B that capture images of predetermined points in a detection target area at different angles of view, and an image detection function unit that detects target images included in image data captured by these cameras 1A and 1B. 2, and a correlation detection function unit 3 that detects a correlation between positions in the detection target area of target images respectively detected by the image detection function unit 2 from image data taken by one camera 1A and another camera 1B; A spoofing determination function unit 4 that determines whether or not the target image in the image data of the one camera 1A and the other camera 1B is a spoofed image from the target image having the correlation detected by the correlation detection function unit 3; and has.

According to the determination device according to the present invention configured as described above, images taken by one camera and another camera are compared between correlated images in a common target area, and the comparison result is It is possible to determine the presence or absence of spoofing.

An image determination method according to a minimum configuration example of the present invention will be described with reference to FIG. 2.
This image determination method includes a step SP1 in which a predetermined point in a detection target area is photographed using a plurality of cameras having different angles of view, a target image detection step SP2 in which a target image included in the photographed image data is detected; A step SP3 of detecting a correlation between the positions of target images detected in the detection target area from image data taken by the camera and another camera, and detecting a correlation between the positions of the target images detected by the camera and the other camera from the target images having the detected correlation; and a step SP4 of determining whether or not the target image in the image data is a spoofed image.

According to the determination method according to the present invention configured as described above, images taken by one camera and another camera are compared with each other in a common target area, and the comparison results are calculated. It is possible to determine the presence or absence of spoofing.

A configuration according to an embodiment of the present invention that embodies FIGS. 1 and 2 will be described with reference to FIGS. 3 to 12.
FIG. 3 shows the overall configuration of a passage determination device according to an embodiment.
The face authentication device 100 receives input of image data from a camera 200 and a camera 201 that shoot in opposite directions. The face recognition device 100 is also connected to a door-shaped gate 300 that opens and closes a traffic lane formed between the side walls on both sides, and information on the opening/closing state of the gate 300 is inputted. It outputs a signal for operating the gate 300 and/or locking and unlocking the gate 300 (locking and unlocking).
The walk-through gate controls traffic in both left and right directions in the figure, and one of the cameras 200 and 201 always photographs the point of the gate 300 in the entrance direction, and the other camera photographs in the exit direction. are doing. That is, the cameras 200 and 201 photograph users in the entrance lane and the exit lane, respectively, in the direction intersecting the gate 300, in order to authenticate the person in both directions of entry and exit.
When the authentication device 100 detects the human body and/or face of the subject attempting to pass through the gate 300, the authentication device 100 detects the detection results from both cameras 200 and 201 and the gate 300, which has been calculated and correlated in advance. Based on the coordinate information of the position corresponding to , it is determined whether the target person having the detected human body and face was present at the detected location. The details of the determination process will be described later.

FIG. 4 is a block diagram showing the configuration of the face authentication device 100.
Face authentication software 110, 120 installed in this face authentication device 100 receives image data from the cameras 200, 201 by image acquisition function units 111, 121, respectively, and receives image data from face/human body detection function units 112, 122. , the position correlation detection function units 113 and 123, and the detected information synthesis function units 115 and 125.
Further, the face recognition software 110, 120 detects the reference data regarding the position of the image, which is detected by the positional correlation detection function units 113, 123 by referring to data regarding the reference correlation in the comparison image databases 114, 124. Correlation values are calculated by comparison and supplied to the detection information synthesis function units 115 and 125 for comprehensive evaluation along with image data judgment.Based on the evaluation results, the impersonation judgment function units 116 and 126 determine whether or not there is impersonation. Based on these determination results, the comprehensive determination function units 117 and 127 determine whether or not the vehicle can pass through the gate. Based on the determination results of these two comprehensive determination function units 117 and 127, for example, on the condition that both comprehensive determination function units 117 and 127 have issued a determination result that there is no impersonation (passage is allowed), the solution is determined. The lock determination function unit 130 determines that the lock can be unlocked, unlocks the gate 300, and operates an actuator or the like to open the gate.

The operation of the face authentication device having the above configuration will be explained together with the steps of the face authentication method.
As shown in FIG. 5, the face authentication device 100 performs, as an initial procedure, a process of setting basic data for correlating the positions of the detection results of the two face authentication software 110 and 120.
The face authentication software 110 uses the image acquisition function section 111 to acquire image data taken by the camera 200. Further, after detecting a reference subject, the face/human body detection function unit 112 in the face authentication software 110 converts the captured image inputted by the positional correlation detection function unit 113 into a rectangle of a predetermined size (for example, , 4 pixels x 4 pixels), and information on which rectangle the detected person belongs to is stored in the comparative image database 114. Similarly, the face authentication software 120 uses the image acquisition function unit 121 to acquire image data taken by the camera 201 from a direction opposite to the camera 200 . After detecting a reference subject, the face/human body detection function section 122 decomposes the current image inputted by the position correlation detection function section 123 into rectangles of a predetermined size (for example, 4 pixels x 4 pixels), Information about which rectangle the detected person belongs to is stored in the comparison image database 124.

In this way, both face recognition software 110 and 120 perform a process of decomposing the image information into rectangles and analyzing which rectangle the detected person is in, and then save information about which rectangle the detected person belongs to. After that, as shown in FIG. 6, the detection information is transmitted from the position correlation detection function section 113 to the position correlation detection function section 123. As shown in FIG. 11, the positional correlation detection function unit 123 associates each rectangle between the image a of the camera 200 and the image b of the camera 201, and then uses the result of the association to perform the positional correlation detection. It is returned to the functional unit 113 and simultaneously stored in the comparative image database 124. The positional correlation detection function unit 113 stores the received association results in the comparison image database 114. That is, both comparison image databases 114 and 124 store the results of association of each rectangle.

For example, assuming that the cameras 200 and 201 acquire image data photographed from opposite directions (directions different by 180 degrees), the correlation performed by the positional correlation detection function unit 123 is as shown in FIGS. 11(a) and 11(b). As shown in , it becomes an image with a bilaterally symmetrical contour shape (the silhouette is bilaterally symmetrical), and it refers to the association of rectangles in a corresponding positional relationship (existing on the same coordinates) in this bilaterally symmetrical image. Specifically, as shown in FIG. 11, the outer edge of the gate in one image (a) corresponds to the outer edge of the other image (b), and the inner edge of the gate in one image (a) corresponds to the outer edge of the gate in one image (a). This is done by linking that the edge of the image corresponds to the inner edge of the other image (b). Also regarding the area (area specified by predetermined coordinates around the gates) where images of various parts of the body of passersby between the gates are captured (at least the head, which is easy to impersonate based on facial recognition). Similarly, the images captured by the left and right cameras 200 and 201 are associated with each rectangle.

After the initial procedure of the above steps is completed, the face authentication device 100 temporarily enters a standby state as shown in FIG. Wait for the data to be output (wait for the image to change as the passerby is photographed).
When a passerby passes by the walk-through gate in this waiting state, the camera 201 on the side facing the passerby takes an image, and the face/human body detection function unit uses the image data acquired by the image acquisition function unit 121. 122 detects a face and a person, as shown in FIG. 8, the result is sent to the comprehensive determination function section 127 and the positional correlation detection function section 123.

The position correlation detection function section 113 acquires the position of the detected person based on the data stored in the comparison image database 124. On the other hand, when a passerby enters the walk-through gate and the passerby is reflected in the camera 200 behind the passerby, the face/human body detection function detects the passerby from the image data acquired by the image acquisition function unit 111. The unit 112 detects a person and transmits the result to the position correlation detection function unit 113.

As shown in FIG. 8, the position correlation detection function section 113 acquires the position of the detected person based on the data of the comparison image database 114, and transmits it to the position correlation detection function section 123. As shown in FIG. 9, the other position correlation detection function unit 123 that has received data from one position correlation detection function unit 113 sends the detected information synthesis function unit 125 to the one position correlation function unit 113 and the other position correlation function unit 113. The results detected by the functional unit 123 are transmitted. As shown in FIG. 12, the detection information synthesis function unit 125 associates, for example, the temporal regions of the passerby's outer gate side in images (a) and (b), and The received data are associated with each other so that the data are associated with each other, and the data is transmitted to the impersonation determination function unit 126.

As shown in FIG. 10, based on the received data, the impersonation determination function unit 126 determines whether there is a real person (for example, a The unlocking determination function determines whether the determination results analyzed for each of the rectangles (a) and (b) in one direction and the opposite direction, which are two-dimensionally expressed, do not match. 130 is notified of the determination result as to whether or not it is an impersonation. At the same time, the verification result from the comprehensive determination function section 127 is received, and based on the information, the unlock determination function section 130 makes a final determination as to whether or not the lock can be unlocked.If the verification is successful and the lock is real, the The gate 300 is controlled to perform operations such as opening and unlocking the gate, and in other cases, a warning display is displayed without opening or unlocking the gate.
In addition to the above-mentioned spoofing judgment, in actual gate operation, facial features such as contours, relative relationships between eyes, ears, nose, mouth, etc. are determined from images taken by one camera (shot from the front). The opening/closing and locking/unlocking of gates is controlled based on the results of facial recognition, which compares the commonality of features with reference data to determine identity.

In this way, in one embodiment, the equipment that photographs passersby in the entrance direction and the exit direction and performs authentication based on the photographed images is used as is, and the images of one camera and the other camera are By determining whether images of corresponding portions match, it is possible to determine whether or not impersonation has occurred, without adding special image acquisition equipment for impersonation determination.

Although one embodiment of the present invention has been described above in detail with reference to the drawings, the specific configuration is not limited to this embodiment, and may include design changes without departing from the gist of the present invention.

INDUSTRIAL APPLICATION This invention can be utilized for the image determination apparatus, traffic management apparatus, and image determination method in entrance-exit management etc.

1A, 1B camera 2 image acquisition function section 3 correlation detection function section,
4 Impersonation determination function unit 100 Face authentication device 110, 120 Face recognition software 111, 121 Image acquisition function unit 112, 122 Face/human body detection function unit 113, 123 Position correlation detection function unit 114, 124 Comparison image database 115, 125 Detection information Synthesis function section 116, 126 (Impersonation) judgment function section 117, 127 Comprehensive judgment function setting section 130 Unlock judgment function section 200, 201 Camera 300 Gate

Claims (6)

a plurality of cameras that photograph predetermined points in the detection target area at different angles of view;
The target image included in the image data taken by these cameras is sent to the image acquisition function unit,
a correlation detection function unit that detects a correlation between positions in the detection target area of target images respectively detected by the image acquisition function unit from image data taken by one camera and another camera;
a spoofing determination function unit that determines whether the target image in the image data of the one camera and the other camera is a spoofed image from the target image having the correlation detected by the correlation detection function unit;
An image determination device having: the plurality of cameras photograph the predetermined point from mutually opposite directions;
The image determination device according to claim 1. The correlation detection function unit decomposes image data of a subject captured by the plurality of cameras into a plurality of target images, respectively;
The spoofing determination function unit determines whether or not the subject is real by comparing the target images at the points of each decomposed part.
The image determination device according to claim 1 or 2. The face/human body detection function unit decomposes the image data into rectangles of a predetermined size and detects which rectangle the target image belongs to,
The correlation detection function unit associates and stores a rectangle between an image of one camera and an image of another camera.
The image determination device according to any one of claims 1 to 3. The image determination device according to any one of claims 1 to 4,
a gate device that restricts the passage of people in the area photographed by the camera;
an unlocking determination unit that releases the traffic restriction by the gate device based on the determination result of the image determination device;
A traffic management device equipped with a step of photographing a predetermined point in the detection target area using a plurality of cameras with mutually different angles of view;
a target image detection step of acquiring a target image included in the photographed image data;
detecting the correlation between the positions of the target images detected by one camera and the other camera in the detection target area;
determining whether or not the target image in the image data of the one camera and the other camera is a spoofed image from the target image having the detected correlation;
An image determination method having the following.

Images