JP2023007006A - Face authentication system and face authentication method - Google Patents

Abstract

To provide a face authentication system and a face authentication method that are capable of preventing impersonation using a face photograph or a display image.SOLUTION: A face authentication system acquires a first image G1 and a second image G2 obtained by imaging an object which is the same as the object of the first image G1 with a changed exposure value. When a difference between brightness L1 of a background image of the first image G1 and brightness L2 of a background image of the second image G2 is a predetermined value or more, the face authentication system determines that the images are images obtained by imaging a person and there are no impersonation, and performs face authentication.SELECTED DRAWING: Figure 5

Description

The present disclosure relates to face authentication systems and face authentication methods.

Japanese Patent Application Laid-Open No. 2008-139973 (Patent Document 1) discloses a face image authentication system that performs personal authentication using an image captured by a camera. This facial image recognition system detects the brightness of the whites of the eyes from the image of the person to be authenticated, and adjusts either the camera exposure or the amount of flash light based on the detected brightness of the whites of the eyes to image the person to be authenticated. Then, the captured image is used for authentication.

JP 2008-139973 A

In a face authentication system using an image, there is a concern that a printed photograph of the face or an image published on the web (World Wide Web) may be presented to a camera to impersonate another person. As disclosed in Patent Document 1, based on the brightness of the white part of the eye, either the exposure of the camera or the light emission amount of the flash is adjusted, and even if an image is taken of a face photograph or a tablet (display) image, There is a possibility that face authentication will be established and spoofing cannot be prevented.

An object of the present disclosure is to provide a face authentication system and a face authentication method capable of deterring spoofing using a face photograph or display image.

A face authentication system according to the present disclosure is a face authentication system that performs authentication by matching a face image captured by a camera with a reference face image. The face authentication system includes a camera control unit that controls a camera to capture a first image and a second image with different exposure-related parameters, focusing on the same subject, the first image and the second image. a background area detection unit that identifies the background, which is the peripheral portion of the face area, as a background image in the two images; a spoofing determination unit that determines spoofing using the background image of the first image and the background image of the second image; Prepare.

According to this face authentication system, the first image and the second image are images of the same subject, focused on the subject, and captured with different parameters related to exposure. When the subject is a real person, the face area and the background around the face area are at different distances from the camera. The distance from the camera to the background, which is the periphery of the face area, is longer (further) than the distance from the camera to the face area. Therefore, in the first image and the second image having different exposure-related parameters, the brightness and degree of blurring of the background around the face region in the first image are the same as the brightness of the background around the face region in the second image. It differs greatly from the degree of bokeh.

A photograph of a face or an image of a display is two-dimensional (flat), and a face area and a background surrounding the face area in the photograph of the face or the image of the display are on the same plane and are at the same distance from the camera. Since the distance from the camera is the same between the face area and the background around the face area in the photograph of the face or the display image, the face in the first image and the second image with different parameters related to exposure are different from each other. The brightness and degree of blurring of the background around the region are not significantly different from the brightness and degree of blurring of the background around the face region of the second image.

The background region detection unit identifies the background, which is the peripheral portion of the first image and the second image at a predetermined distance from the face region, as the background image, and the spoofing determination unit detects the difference between the background image of the first image and the second image. A background image is used to determine spoofing. For example, if the brightness and the degree of blurring of the background images of the first image and the second image hardly change, there is a high possibility that the background image of the first image and the second image are spoofed by a photograph of a person's face or an image of the display. It becomes possible to judge spoofing by using the background image of . This makes it possible to suppress spoofing without providing a dedicated device for capturing the features of facial photographs and display images.

The camera control unit captures the first image with a predetermined exposure, captures the second image with illumination brighter than the illumination at the time of capturing the first image, and with an exposure value greater than the predetermined exposure, and the spoofing determination unit captures the first image. When the difference in brightness between the background image of the second image and the background image of the second image is equal to or greater than a predetermined value, it may be determined that the image is not spoofing.

When the subject is a real person, the distance from the camera to the background, which is the periphery of the face area, is longer than the distance from the camera to the face area. For this reason, if the exposure value of the camera is increased and the subject is imaged by illuminating it, the brightness of the background around the face region will be lower and darker than when the exposure value is small. When the subject is a face photograph or display image, the face area and the background around the face area in the face photograph or display image are on the same plane and are at the same distance from the camera. Therefore, even if the exposure value of the camera is increased and the subject is illuminated and captured, the brightness of the background around the face region does not change much compared to when the exposure value is small.

Therefore, when the difference in brightness between the background image of the first image and the background image of the second image is equal to or greater than a predetermined value, it can be determined that the image is of a real person and not impersonation. Note that the exposure value (EV) is a value that adjusts the amount of light reaching the imaging element of the camera, and in the present disclosure, the smaller the amount of light reaching the imaging element, the larger the value. The exposure value may be set from at least one of the aperture value (F number), shutter speed (exposure time), and ISO (International Organization for Standardization) sensitivity of the camera, or a combination thereof.

The camera control unit captures a first image with a predetermined exposure, captures a second image with illumination brighter than the illumination when the first image was captured, and an exposure value greater than the predetermined exposure, and the spoofing determination unit captures the first When the difference in brightness between the image of the face area of the image and the background image is smaller than a predetermined value and the difference in brightness between the image of the face area of the second image and the background image is equal to or greater than a predetermined value, it is determined that the image is not spoofing. You may do so.

When the subject is a real person, the distance from the camera to the background, which is the periphery of the face area, is longer than the distance from the camera to the face area. For this reason, if the exposure value of the camera is increased and the subject is illuminated and captured, the brightness of the background surrounding the face area will decrease and become darker than when the exposure value is small. , since it is closer to the camera than the background, which is the periphery of the face area, the change in luminance due to the difference in exposure value is smaller than that of the background. When the subject is a face photograph or display image, the face area and the background around the face area in the face photograph or display image are on the same plane and are at the same distance from the camera. Therefore, even if the exposure value of the camera is increased and the subject is illuminated and captured, the brightness of the face area and the brightness of the background surrounding the face area will not change much compared to when the exposure value is small. Absent.

Therefore, when the difference in brightness between the image of the face region in the first image and the background image is smaller than a predetermined value and the difference in brightness between the image of the face region in the second image and the background image is greater than or equal to the predetermined value, A real person is imaged, and it can be determined that it is not impersonation.

The camera control unit captures the first image with a predetermined aperture value, and captures the second image with an aperture value smaller than the predetermined aperture value. It may be determined that the image is not spoofing when the difference in the degree of blurring of the image is equal to or greater than a predetermined value.

When the subject is a real person, the distance from the camera to the background, which is the periphery of the face area, is longer than the distance from the camera to the face area. For this reason, if the aperture value of the camera is made small and the depth of field is made shallow, the degree of blurring of the background around the face area (bokeh) will be lower than when the aperture value is large and the depth of field is large. amount) increases. When the subject is a face photograph or display image, the face area and the background around the face area in the face photograph or display image are on the same plane and are at the same distance from the camera. For this reason, even if the depth of field changes by changing the aperture value of the camera, the photograph of the face and the image on the display are in focus, so the background blur around the face area in the photograph of the face and the image on the display is blurred. The degree (bokeh amount) does not increase.

Therefore, when the difference in degree of blur between the background image of the first image and the background image of the second image is equal to or greater than a predetermined value, it can be determined that the image is not spoofing.

The authentication system of the present disclosure further includes a face extraction unit that extracts a face image from an image captured by a camera, and a matching determination unit that matches the face image with a reference face image to perform face authentication. may perform face authentication by collating a face image extracted from at least one of the first image and the second image with a reference face image when the spoofing determination unit determines that the person is not spoofing.

According to this configuration, face authentication is performed when it is determined that the image captured by the camera is not spoofed. Therefore, since the face image to be compared with the reference face image is not the photograph of the face or the image of the display, it is possible to suppress the establishment of authentication by spoofing using the photograph of the face or the image of the display.

The face authentication method of the present disclosure is a face authentication method that performs authentication by matching a face image captured by a camera with a reference face image. The face authentication method includes the steps of: focusing on the same subject and capturing a first image and a second image with different parameters related to exposure with the camera; and The method includes a step of identifying a background that is a peripheral portion of the face area as a background image, and a step of determining spoofing using the background image of the first image and the background image of the second image.

According to this face authentication method, the first image and the second image are images of the same subject, focused on the subject, and captured with different parameters related to exposure. When the subject is a real person, the face area and the background around the face area are at different distances from the camera. The distance from the camera to the background, which is the periphery of the face area, is longer than the distance from the camera to the face area. Therefore, between the first image and the second image captured with different parameters related to exposure, the difference in brightness and degree of blurring of the background around the face area is large.

A photograph of a face or an image of a display is two-dimensional (flat), and a face area and a background surrounding the face area in the photograph of the face or the image of the display are on the same plane and are at the same distance from the camera. Since the distance from the camera is the same between the face area and the background around the face area in the photograph of the face or the image of the display, the first image and the second image captured with different parameters related to exposure show the face The difference in brightness and degree of blurring of the background, which is the periphery of the region, is small.

Therefore, for example, when the brightness and the degree of blurring of the background images of the first image and the second image hardly change, there is a high possibility that the background image of the first image and the background image of the second image are spoofed by a photograph of a person's face or an image of the display. Spoofing can be determined by using two background images.

Advantageous Effects of Invention According to the present disclosure, it is possible to provide a face authentication system and a face authentication method capable of deterring spoofing using a photograph of the face or an image on a display.

1 is a schematic configuration diagram of a vehicle 1 for car sharing equipped with a face authentication system of the present disclosure; FIG. 3 is a diagram showing functional blocks configured within an authentication ECU 200. FIG. (A) and (B) are diagrams showing images of a person captured by the camera 10. FIG. (A) and (B) are diagrams showing images of a printed facial photograph p captured by the camera 10. FIG. 4 is a schematic flowchart of processing executed by an authentication ECU 200; FIG. 7 is a functional block diagram configured in an authentication ECU 200a in Embodiment 2; 4 is a schematic flowchart of processing executed by an authentication ECU 200a; 10A and 10B are diagrams showing images of a person captured by the camera 10 in Embodiment 3. FIG. 12A and 12B are diagrams showing images of a printed facial photograph p captured by the camera 10 in the third embodiment. FIG. FIG. 11 is a functional block diagram configured in an authentication ECU 200b in Embodiment 3; 4 is a schematic flowchart of processing executed by an authentication ECU 200b;

Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. The same or corresponding parts in the drawings are denoted by the same reference numerals, and the description thereof will not be repeated.

(Embodiment 1)
In this embodiment, an example in which the face authentication system of the present disclosure is applied to car sharing will be described. FIG. 1 is a schematic configuration diagram of a vehicle 1 for car sharing equipped with a face authentication system of the present disclosure. Car sharing is a mobile service in which registered members jointly use a vehicle 1 . A vehicle 1 is provided with a door 3 on a vehicle body 2 . A user of the vehicle 1 can get on the vehicle 1 through the door 3 and use the vehicle 1 . The vehicle 1 is provided with a door lock device 30, which locks and unlocks the door 3. As shown in FIG. The vehicle 1 is arranged in a rental-only parking lot, and a person who wishes to use the vehicle 1 makes a reservation for use, goes to the parking lot on the reserved date and time, unlocks the door, and uses the vehicle 1.例文帳に追加

A vehicle 1 is equipped with a face authentication system 5. - 特許庁The face authentication system 5 includes a camera 10 , an electronic flash 20 , a camera ECU (Electronic Control Unit) 100 , an authentication ECU 200 and a communication device 250 .

The camera 10 captures an image of a subject around the door 3 to acquire image data. The camera 10 includes, for example, an imaging element such as a CCD (Charge Coupled Device) image sensor, a CMOS (Complementary Metal Oxide Semiconductor) image sensor, and an optical system (lens, diaphragm, shutter) that forms an image of a subject on the imaging element. etc. is a digital camera.

The electronic flash 20 is a light-emitting device that irradiates a subject with illumination (flash light), and is provided near the imaging lens of the camera 10 .

The camera ECU 100 includes a CPU (Central Processing Unit) and memory (for example, ROM (Read Only Memory) and RAM (Random Access Memory)). The camera ECU 100 receives signals and instructions from the authentication ECU 200 and controls operations of the camera 10 and the electronic flash 20 . The camera ECU 100 includes a focus control section 101, an exposure control section 102, and a light emission control section 103 as its functional blocks. The focus control unit 101 detects the positions of the eyes, nose, and mouth of a subject (person) within the imaging range of the camera 10, detects the person's face based on the position information, and detects the face detected by the camera 10. to focus on and perform autofocus processing. The exposure control unit 102 adjusts the aperture value (F value), shutter speed (exposure time), and ISO sensitivity when capturing an image with the camera 10 . The light emission control unit 103 controls light emission/non-light emission of the electronic flash 20 and adjusts the amount of light during light emission.

The face authentication system 5 performs face authentication by matching a face image detected from an image captured by the camera 10 controlled by the camera ECU 100 with a reference face image. A reference face image is a face image of a registered member (user) and is registered in the management server 400 . The management server 400 includes an ECU 401 , a storage device 402 and a communication device 403 . ECU 401 includes a CPU and is configured to perform predetermined information processing and control communication device 403 . The storage device 402 is configured to be able to store various types of information. A communication device 403 includes various communication I/Fs. The ECU 401 is configured to be able to communicate with the outside through the communication device 403, and communicates with the communication device 250 of the vehicle 1 and the mobile terminal 500 of the member (user).

A member (user) registers member information in advance. When member information is registered, the ID (identification code) of the member and face image data of the member are registered in the storage device 402 of the management server 400 . The member's face image data is obtained by transmitting face image (face area image) data taken by the member with a camera (for example, the camera of the mobile terminal 500) to the management server 400, and the ECU 401 extracts facial features from the received face image. is created, and for each member, the ID of the member and the feature amount data Ri are linked and stored in the storage device 402 . The feature amount data Ri representing facial features is extracted from the face image using a machine learning algorithm (for example, deep learning using a convolutional neural network (CNN)). do. It should be noted that feature amount data Ri representing feature amounts of facial parts (eyes, nose, mouth, outline, etc.) may be created from the face image, or data extracted from the entire image of the face region. may The feature amount data Ri stored in the storage device 402 corresponds to the "reference face image" of the present disclosure.

A member (user) uses the mobile terminal 500 to make a reservation (rental reservation) for use of car sharing. For example, when the application of the user's portable terminal 500 is launched, the rental location, rental date and time, and return date and time are entered and the reservation button is tapped, the member's ID, rental location, rental date and time, and return date and time are sent to the management server 400. be done. When the management server 400 receives the reservation information from the mobile terminal 500, the ECU 401 updates the rental reservation DB stored in the storage device 402, and the reservation is completed. The rental location is the location of the parking lot where the vehicle 1 for car sharing is parked.

The member goes to the rental place at the rental date and time, performs face authentication by the face authentication system 5 mounted on the vehicle 1 for car sharing, and when the authentication is established, the door is unlocked and the vehicle 1 can be used. become. Face authentication by the face authentication system 5 is performed by matching a face image detected from an image captured by the camera 10 with a reference face image. At this time, a third party other than the member who made the rental reservation may print the photo of the member who made the rental reservation or the image of the tablet (display) showing the face of the member who made the rental reservation. , is presented to the camera 10, and there is concern that a third party may fraudulently pretend to be a member.

A photo of the face of the member who made the rental reservation or a face image detected from an image of the display showing the face of the member who made the rental reservation captured by the camera 10 is There is a high possibility that the face image matches the reference face image of the selected member, and it is desirable to suppress spoofing using the face photo or display image.

In the present embodiment, in order to suppress this spoofing, the camera 10 captures the first image and the second image with different parameters related to exposure, so that the image captured by the camera 10 can be used as a facial photograph or a display. It is determined that the image is obtained by capturing the image of .

Authentication ECU 200, like camera ECU 100, comprises a CPU and a memory (for example, ROM and RAM). Authentication ECU 200 outputs an imaging command to camera ECU 100 and performs face authentication using an image captured by camera 10 . Details of the authentication ECU 200 will be described later.

The communication device 250 includes various communication I/Fs (interfaces). The communication device 250 may include a DCM (Data Communication Module). The communication device 250 may include a communication I/F compatible with 5G (fifth generation mobile communication system). Authentication ECU 200 is configured to perform wireless communication with management server 400 outside vehicle 1 through communication device 250 .

The door ECU 300 receives a command from the authentication ECU 200 and drives the door lock device 30 to lock and unlock the door 3 .

FIG. 2 is a diagram showing functional blocks configured in the authentication ECU 200. As shown in FIG. A camera control unit 201 controls the camera 10 via the camera ECU 100 . For example, when the mobile terminal 500 is held over a reading unit of an NFC (Near field communication) device installed near the door 3 of the vehicle 1, the camera ECU 100 is instructed to capture a first image and a second image with the camera 10. Outputs an imaging command. The imaging command is to image the subject with proper exposure (automatic exposure) as the first image, emit the electronic flash 20 as the second image, and set the exposure value to be larger than the proper exposure when the first image was captured. This is a command to image an object. The imaging of the first image and the imaging of the second image are continuously performed, and the first image and the second image are images of the same subject.

When the camera ECU 100 receives an imaging command, the focus control unit 101 focuses on the face area of the subject (person), and the exposure control unit 102 adjusts the aperture value (F value), shutter speed (exposure time), and so on to achieve proper exposure. ), and adjust the ISO sensitivity to capture the first image. The camera 10 outputs the captured image as the first image G1. After the camera ECU 100 captures the first image, the light emission control unit 103 causes the electronic flash 20 to emit light with a predetermined amount of light emission. An area is focused, and at least one of the aperture value (F value), shutter speed (exposure time), and ISO sensitivity is adjusted by the exposure control unit 102 so that the exposure value becomes larger than the exposure value when the first image was captured. to operate the camera 10 to capture a second image. The camera 10 outputs the captured image as the second image G2.

The exposure value at the time of capturing the second image may be, for example, an exposure value that provides proper exposure at the distance of the face region when the light (illumination) emitted by the electronic flash 20 is applied to the face region of the subject. . As a result, the subject's face area is brightened by the emission of the electronic flash 20, so that the exposure value is increased (the amount of light reaching the imaging device of the camera 10 is reduced) compared to when the electronic flash 20 is not emitted. To increase the exposure value, the aperture value (F number) of the camera 10 may be increased, the shutter speed may be increased (exposure time shortened), and the ISO sensitivity may be decreased. May be combined.

FIG. 3 is a diagram showing an image of a person captured by the camera 10. As shown in FIG. FIG. 3A shows the first image G1, and FIG. 3B shows the second image G2. In the first image G1 captured without the electronic flash 20 emitting light, as shown in FIG. 3A, both the background b represented by the lattice pattern and the person's face f are captured with proper brightness. In the second image G2 captured with the electronic flash 20 emitted and the exposure value increased, as shown in FIG. Since the distance from the camera 10 is longer than the face f and the light from the electronic flash 20 is difficult to reach, the image is dark.

FIG. 4 is a diagram showing an image of a printed portrait p taken by the camera 10. As shown in FIG. FIG. 4A shows the first image G1, and FIG. 4B shows the second image G2. In the first image G1 captured without the electronic flash 20 emitting light, as shown in FIG. The face f is also clearly imaged. In the second image G2 captured by emitting the electronic flash 20 and increasing the exposure value, as shown in FIG. Since it is difficult for the light from the flash 20 to reach the image, the image is dark. In this case, as shown in FIG. 4B, the hatched background bp and the face f in the facial photograph p are on the same plane, and the distance from the camera 10 is the same. f is sharply imaged with similar brightness.

When the camera 10 captures an image of a person, the second image G2 has a darker background b than the first image G1, as can be understood from the comparison between FIGS. 3A and 3B. On the other hand, when the printed facial photograph p is captured, as understood from the comparison between FIGS. 4A and 4B, the brightness of the background bp of the facial photograph is There is almost no change in one image G1. Therefore, by comparing the brightness of the background, which is the peripheral portion of the face f, in the first image G1 and the second image G2, if the difference in brightness is equal to or greater than a predetermined value, the image is an image of a person, and the brightness is low. If the difference in height is smaller than a predetermined value, it can be determined that the image is a face photograph.

Referring to FIG. 2 , authentication ECU 200 includes face/background area detection section 202 . The face/background area detection unit 202 uses a machine learning algorithm to detect and identify the face area from the first image G1 and the second image G2 captured by the camera 10 . Note that the positions of the eyes, nose, and mouth of the subject may be detected, and based on the position information, the face area of the image may be detected and the face area may be specified. A range (margin range) set in advance in consideration of a person's hairstyle is defined as a hair region around the identified face region. Then, a predetermined range around the hair region is identified as the background region. The predetermined range may be a range within a predetermined distance from the hair region. Also, the predetermined range may be defined by the number of pixels in the X direction and the Y direction in the image around the hair region. The number of pixels may be, for example, 1/10 of the total number of pixels in the X direction and 1/10 of the total number of pixels in the Y direction. If the hairstyle can be identified from the RGB (Red, Green, Blue) information of the first image and the second image, a predetermined range around the hairstyle may be identified as the background area. The background area specified in this way is the area for discriminating between the background b when a person is imaged and the background bp of the face photo p when a printed face photo p is imaged.

The face/background area detection unit 202 calculates the background image brightness L1 from the background area of the first image G1, and calculates the background image brightness L2 from the background area of the second image G2. For example, the relationship between the RGB gradation and luminance of each pixel included in the background area of the image is obtained in advance, and the luminance of each pixel is calculated from the RGB gradation of each pixel. Then, the calculated average value of the brightness of each pixel may be calculated as the brightnesses L1 and L2 of the background image.

The spoofing determination unit 203 compares the background image brightness L1 of the first image G1 and the background image brightness L2 of the second image G2 calculated by the face/background area detection unit 202, and captures the image with the camera 10. It is determined whether or not the image is a spoofed image using a photograph of the face. If the difference between the brightness L1 and the brightness L2 (luminance difference) is equal to or greater than a predetermined value, the image is determined to be an image of a person and is not spoofing, and a signal indicating that the image captured by the camera 10 is not spoofing is generated. Output. If the difference between the brightness L1 and the brightness L2 (luminance difference) is smaller than a predetermined value, the image is determined to be a photograph of a person's face and is a spoofing image. Note that the predetermined value is set in advance by experiment or the like.

When the face extraction unit 204 receives a signal from the spoofing determination unit 203 indicating that the image captured by the camera 10 is not spoofing, the face extraction unit 204 extracts a face feature amount (face feature amount) from the first image G1 or the second image G2. do. For example, from the image of the face region of the first image G1 identified by the face/background region detection unit 202, using a machine learning algorithm (for example, deep learning using a convolutional neural network (CNN)), Extract the feature amount data Fi. Note that feature amount data Fi representing feature amounts of facial parts (eyes, nose, mouth, outline, etc.) may be created from the image of the face region. The feature amount data Fi extracted by the face extraction unit 204 corresponds to the “face image” of the present disclosure, and is hereinafter also referred to as the face image. The feature amount data Fi may be extracted from the face area of the second image G2. The feature amount data Fi (face image) may be either data extracted from a part of the face part or data extracted from the entire image of the face region.

A reference face image acquisition unit 205 acquires a reference face image (feature amount data Ri) from the management server 400 . For example, when the member completes the rental reservation of the vehicle 1 , the management server 400 reads from the storage device 402 the reference face image (feature data Ri) associated with the ID of the member who made the reservation of the vehicle 1 . Then, the management server 400 uses the communication device 403 to link and transmit the rental date and time, the member's ID, and the reference face image to the vehicle 1 . Upon receiving the rental date and time, the member's ID, and the reference facial image via the communication device 250, the reference facial image acquisition unit 205 associates the rental date and time, the member's ID, and the reference facial image with the memory of the authentication ECU 200. memorize.

The matching determination unit 206 performs face authentication by matching the face image (feature amount data Fi) extracted by the face extraction unit 204 with the reference face image (feature amount data Ri). For example, the reference face image associated with the closest rental date and time to the date and time when the first image G1 or the second image G2 was captured is read out from the memory. Then, the face image detected by the face extraction unit 204 is compared with the reference face image read from the memory to perform face authentication. The matching determination unit 206 compares, for example, the feature amount data Ri of the reference face image with the feature amount data Fi of the face image detected by the face extraction unit 204, and determines that the similarity between the two feature amount data is a predetermined value. If they match, it is determined that they match (face authentication is successful), and if the degree of similarity is less than a predetermined value, it is determined that they do not match (face authentication is not successful).

When the matching determination unit 206 determines that the feature amount data Ri of the reference face image and the feature amount data Fi of the face image match and the face authentication is established, it issues an unlock command to the door ECU 300 to unlock the door. to output Upon receiving the unlock command, the door ECU 300 drives the door lock device 30 to unlock the door 3 .

FIG. 5 is a schematic flow chart of the process executed by authentication ECU 200. As shown in FIG. This flowchart is executed when the portable terminal 500 is held over the reader of the NFC device installed near the door 3 of the vehicle 1 . First, in step (hereinafter, step is abbreviated as S) 10, the camera 10 captures an image of a subject to obtain a first image G1. For the first image G1, as described above, the focus control unit 101 focuses on the face area of the subject (person), and the exposure control unit 102 adjusts the aperture value (F value) and shutter speed so that the exposure is appropriate. (Exposure time) and ISO sensitivity are adjusted to capture images.

Subsequently, in S11, the camera 10 captures an image of the subject to acquire the second image G2. As described above, the second image G2 is generated by the light emission control unit 103 causing the electronic flash 20 to emit light with a predetermined amount of light emission, and in synchronization with the light emission, the focus control unit 101 detects the face area of the subject (person). , and the exposure control unit 102 adjusts at least one of the aperture value (F value), shutter speed (exposure time), and ISO sensitivity so that the exposure value is greater than the exposure value when the first image was captured. This is the captured image. Note that S10 and S11 are continuously processed, and the first image G1 and the second image G2 are images of the same subject.

In S12, the background image brightness L1 of the first image G1 and the background image brightness L2 of the second image G2 are calculated. From the first image G1 and the second image G2, a machine learning algorithm is used to detect the face area of the image to identify the face area. Then, an area within a predetermined range that is at least a predetermined distance away from the identified face area is identified as a background area. Then, the brightness of each pixel is calculated from the relationship between the RGB gradation and brightness of each pixel included in the background region of the first image G1, and the calculated average value of the brightness of each pixel is taken as the brightness L1 of the background image. calculate. Similarly, the brightness L2 of the background image is calculated from the relationship between the RGB gradation and brightness of each pixel included in the background area of the second image G2.

Subsequently, in S13, it is determined whether or not the difference between the brightness L1 and the brightness L2 is equal to or greater than a predetermined value. If the difference (luminance difference) between the brightness L1 and the brightness L2 is equal to or greater than the predetermined value, the determination is affirmative and the process proceeds to S14. If the difference (luminance difference) between the brightness L1 and the brightness L2 is smaller than the predetermined value, a negative determination is made, and the current routine ends.

In S14, a face image (feature data Fi) is extracted from the first image G1 or the second image G2. For example, from the image of the face area of the first image G1 specified in S12, the feature amount data Fi representing the features of the face is detected using a machine learning algorithm. The feature amount data Fi may be extracted from the face area image of the second image G2.

In S15, the reference face image (feature amount data Ri) associated with the closest rental date and time to the current date and time (that is, the date and time when the first image G1 or the second image G2 was captured) is read from the memory, and the process proceeds to S16. move on. Note that the reference face image is received in advance from the management server 400 and stored in the memory as described above.

Subsequently, in S16, the face image (feature amount data Fi) extracted in S14 and the reference face image (feature amount data Ri) are collated and determined to perform face authentication. As described above, for example, when the similarity between the two feature amount data is equal to or greater than a predetermined value and they match, face authentication is established and an affirmative determination is made, and the process proceeds to S17. Face authentication is not established and a negative determination is made, and this routine ends.

In S17, an unlock command is output to the door ECU 300 to unlock the door, and the current routine ends.

According to the present embodiment, the first image G1 and the second image G2 are images of the same subject captured by the camera 10, focused on the subject, and captured with different parameters related to exposure. . The first image G1 focuses on the face area of the subject, and adjusts exposure-related parameters such as aperture value (F value), shutter speed (exposure time), and ISO sensitivity so as to achieve proper exposure. , is imaged. The second image G2 is synchronized with the emission of the electronic flash 20, focused on the subject's face area, and has parameters related to exposure such that the exposure value is greater than the exposure when the first image G1 was captured. At least one of the aperture value (F value), shutter speed (exposure time), and ISO sensitivity is adjusted to capture the image.

When the camera 10 captures an image of a person, the distance from the camera 10 to the background is longer than the distance from the camera 10 to the person. Therefore, the background image of the second image G2 is darker than the background image of the first image G1. In the present embodiment, when the difference between the background image brightness L1 of the first image G1 and the background image brightness L2 of the second image G2 is equal to or greater than a predetermined value, the image captured by the camera 10 is an image of a person. and it is determined that it is not spoofing.

When the printed face photo p is captured by the camera 10, the background and the face in the face photo p are on the same plane and the distance from the camera 10 is the same. and the brightness of the background image around the face in the first image G1 are almost the same. In the present embodiment, when the difference between the background image brightness L1 of the first image G1 and the background image brightness L2 of the second image G2 is smaller than a predetermined value, the image captured by the camera 10 is a face photograph. It is an image and is determined to be spoofing. In the present embodiment, an example of a printed facial photograph has been described as a form of spoofing, but the same applies to the image on the display.

As described above, according to the present embodiment, spoofing can be determined and spoofing can be suppressed without providing a dedicated device for capturing the features of a face photograph or a display image.

In the present embodiment, the first image G1 is captured and then the second image G2 is continuously captured. good.

In this embodiment, the reference face image acquisition unit 205 receives and acquires the reference face image (feature data Ri) from the management server 400 in advance. However, the reference face image may be obtained after or at the same time that the first image G1 and the second image G2 are captured. For example, simultaneously with the start of the process of the flowchart of FIG. 5, the vehicle ID of vehicle 1 and the current date and time are transmitted to the management server 400, and the reference face image of the member who has reserved the vehicle 1 at the current date and time is requested, A reference face image may be received from the management server 400 . Also, when the mobile terminal 500 is held up to the NFC device, if the member's ID can be acquired from the mobile terminal 500, a transmission request for the reference face image of the acquired member's ID is transmitted to the management server 400, You may receive the reference face image of the member who does.

In the present embodiment, in the flowchart of FIG. 5, the routine ends when a negative determination is made in S13 and when a negative determination is made in S16. However, when negative determinations are made in S13 and S16, information may be transmitted to the management server 400 to the effect that a person other than the member who made the reservation is going to use the vehicle 1. FIG.

In this embodiment, the electronic flash 20 does not emit light when capturing the first image G1, but it may emit light. In this case, it is desirable that the amount of light emitted by the electronic flash 20 when capturing the first image G1 is sufficiently smaller than the amount of light emitted when capturing the second image G2. Also, any configuration may be used as long as it can illuminate the subject, and instead of the electronic flash 20, a lighting fixture such as an LED (Light Emitting Diode) light may be used.

(Embodiment 2)
In the above embodiment, spoofing is determined using the difference between the background image brightness L1 of the first image G1 and the background image brightness L2 of the second image G2. In Embodiment 2, spoofing is performed using the difference between the brightness of the face f in the first image G1 and the brightness L1 of the background image, and the difference between the brightness of the face f in the second image G2 and the brightness L2 of the background image. judge.

When an image of a person is captured by the camera 10, as shown in FIG. The difference in height L1 is small. However, as shown in FIG. 3B, in the second image G2, the face f of the person is captured with appropriate brightness, but the background b is captured dark. The difference in height L2 becomes large.

When the photograph of the face p printed by the camera 10 is captured, the face f and the background bp are captured with the same brightness in the first image G1 and the second image G2, as shown in FIGS. Therefore, the difference between the brightness of the face f in the first image G1 and the background image brightness L1 is small, and similarly, the difference between the brightness of the face f in the second image G2 and the background image brightness L2 is also small.

Therefore, when the difference between the brightness of the face f in the second image G2 and the background image brightness L2 is larger than the difference between the brightness of the face f in the first image G1 and the background image brightness L1 by a predetermined value or more, the person is captured. It can be determined that the image is an image that has been processed.

FIG. 6 is a functional block diagram configured in authentication ECU 200a in the second embodiment. In the second embodiment, the configuration other than the authentication ECU 200a is the same as that of the first embodiment, so description thereof will be omitted. 6, camera control unit 201, face extraction unit 204, reference face image acquisition unit 205, and collation determination unit 206 are the same as authentication ECU 200 of the first embodiment, so description thereof will be omitted.

The face/background area detection unit 202a uses a machine learning algorithm to detect the face area of the image from the first image G1 and the second image G2 captured by the camera 10, and specifies the face area. Note that the positions of the eyes, nose, and mouth of the subject may be detected, and based on the position information, the face area of the image may be detected and the face area may be specified. As in the first embodiment, a predetermined range of areas separated by a predetermined distance or more from the identified face area is identified as a background area. Then, similarly to the first embodiment, the background image brightness L1 is calculated from the background area of the first image G1, and the background image brightness L2 is calculated from the background area of the second image G2.

The face/background area detection unit 202a calculates the average luminance value of each pixel included in the face area of the first image G1 as the brightness FL1 of the face f of the first image G1, and calculates the brightness FL1 of the face f of the first image G1 as the face area of the second image G2. is calculated as the brightness FL2 of the face f in the second image G2. Then, the face/background area detection unit 202a calculates the difference S1 (=|FL1−L1|) between the brightness FL1 of the face f in the first image G1 and the brightness L1 of the background image, and calculates the difference S1 (=|FL1−L1|) A difference S2 (=|FL2−L2|) between the brightness of f and the brightness L2 of the background image is calculated.

The spoofing determination unit 203a determines whether the difference S1 between the brightness FL1 of the face f in the first image G1 and the brightness L1 of the background image is smaller than a predetermined value α (S1<α) and the brightness of the face f in the second image G2. If the condition that the difference S2 between the brightness L2 of the background image and the background image is equal to or greater than a predetermined value α (S2≧α) is satisfied, it is determined that the image is an image of a person and is not spoofing. In addition, the spoofing determination unit 203a determines that the difference S1 between the brightness FL1 of the face f in the first image G1 and the brightness L1 of the background image is smaller than a predetermined value α (S1<α), and the face f in the second image G2 If the difference S2 between the brightness L2 of the background image and the brightness L2 of the background image is smaller than a predetermined value α (S2<α), the image is determined to be a photograph of a person's face and to be impersonation. If none of the conditions are met, the spoofing determination unit 203a does not determine the presence or absence of spoofing. Note that the predetermined value α is set in advance by experiments or the like.

FIG. 7 is a schematic flowchart of processing executed by the authentication ECU 200a. This flowchart eliminates S12 and S13 of the flowchart of FIG. 5 and adds S20 to S23. In FIG. 7, the processes of S10 to S17 are the same as those in the flowchart of FIG. 5, so description thereof will be omitted.

In S20, the difference S1 (=|FL1−L1|) between the brightness FL1 of the face f and the brightness L1 of the background image in the first image G1, and the brightness of the face f and the brightness of the background image in the second image G2 are calculated. A difference S2 (=|FL2-L2|) of L2 is calculated. From the first image G1 and the second image G2 captured by the camera 10, a machine learning algorithm is used to detect and identify the face area of the image. A region of a predetermined range separated by a predetermined distance or more from the identified face region is identified as a background region, the brightness L1 of the background image is calculated from the background region of the first image G1, and the background region of the second image G2 is calculated. , the brightness L2 of the background image is calculated. The average brightness of each pixel included in the face region of the first image G1 is calculated as the brightness FL1 of the face f in the first image G1, and the average brightness of each pixel included in the face region of the second image G2 is calculated. A value is calculated as the brightness FL2 of the face f in the second image G2. Then, the face/background area detection unit 202a calculates the difference S1 between the brightness FL1 of the face f in the first image G1 and the brightness L1 of the background image, and calculates the difference S1 between the brightness FL1 of the face f in the first image G1 and the brightness L1 of the background image. A difference S2 of the height L2 is calculated.

Subsequently, in S21, it is determined whether or not the difference S1 in brightness is smaller than a predetermined value α (S1<α) and the difference S2 in brightness is equal to or greater than a predetermined value α (S2≧α). In the case of an image of a person captured by the camera 10, the difference S1 between the brightness FL1 of the face f in the first image G1 and the brightness L1 of the background image is small, and the brightness of the face f and the brightness of the background image in the second image G2 is small. Since the difference S2 of L2 is large, an affirmative determination is made in S21, and the process proceeds to S14.

If the determination in S21 is negative, the process advances to S22 to determine whether the brightness difference S1 is smaller than a predetermined value α (S1<α) and the brightness difference S2 is smaller than a predetermined value α (S2<α). judge. In the case of the image obtained by capturing the photograph of the face p printed by the camera 10, the difference S1 between the brightness FL1 of the face f in the first image G1 and the brightness L1 of the background image is small, and the brightness of the face f and the background in the second image G2 is small. Since the difference S2 of the brightness L2 of the image is also small, an affirmative determination is made in S22, and the process proceeds to S23.

In S23, it is determined that the image captured by the camera 10 is a spoofed image of the facial photograph p, and information to that effect is transmitted to the management server 400, for example, and the current routine ends.

If a negative determination is made in S22, it cannot be determined whether the image captured by the camera 10 is an image of a person or an image of a photograph of the face p, so the current routine is terminated. .

When a person is captured by the camera 10, the distance from the camera 10 to the background is longer than the distance from the camera 10 to the person. is smaller than the predetermined value α, and the difference S2 between the brightness of the face f in the second image G2 and the brightness L2 of the background image is greater than or equal to the predetermined value α. Therefore, according to the second embodiment, it can be determined that the image captured by the camera 10 is an image of a person and is not impersonation.

When the printed facial photograph p is captured by the camera 10, the background and the face appearing in the facial photograph p are on the same plane, and the distance from the camera 10 is the same. The difference S1 between the brightness FL1 of the face f and the brightness L1 of the background image is less than a predetermined value α, and the difference S2 between the brightness of the face f in the second image G2 and the brightness L2 of the background image is less than a predetermined value α. . Therefore, according to the second embodiment, the image captured by the camera 10 is an image of a photograph of the face, and can be determined to be spoofing.

As described above, according to the second embodiment, spoofing can be determined and spoofing can be suppressed without providing a dedicated device for capturing the features of the face photograph or the image on the display.

(Modification)
In Embodiments 1 and 2 described above, spoofing is determined by comparing the brightness L1 of the background image of the first image G1 and the brightness L2 of the background image of the second image G2. However, spoofing may be determined by comparing the overall brightness of the first image G1 and the second image G2. For example, the luminance difference is calculated for each pixel at the same coordinates in the first image G1 and the second image G2, and the luminance difference distribution is obtained. Then, if the spread of the distribution is greater than or equal to a predetermined value, or if the peaks of the distribution are separated, it is determined that the image is an image of a person and is not impersonation, the spread of the distribution is within a predetermined value, and , if the peaks of the distribution are not separated, it may be determined that the image is a photograph of a face or an image of a display and that the image is spoofing.

In addition to the spoofing determination of the first embodiment or the second embodiment, the spoofing determination of the above-described modified example can be performed to achieve more accurate background b when a person is imaged and a printed facial photograph. It becomes possible to discriminate the background bp of the facial photograph p when p is imaged.

(Embodiment 3)
In Embodiment 1, spoofing is determined using the difference between the background image brightness L1 of the first image G1 and the background image brightness L2 of the second image G2. In the third embodiment, spoofing is determined using the degree of blurring of the background image of the first image and the background image of the second image. In Embodiment 3, the basic configuration of a vehicle equipped with a face authentication system is practically the same as that of vehicle 1 in FIG. 1, and only the configuration of authentication ECU 200 is different.

FIG. 8 is a diagram showing an image of a person captured by the camera 10 in the third embodiment. FIG. 8A shows the first image g1 captured by setting the aperture value (F number) of the camera 10 to a predetermined value (for example, F16), and FIG. For example, the second image g2 captured with F2.8). Note that the camera 10 is focused on the person's face f by autofocus processing. When the subject is a person, the distance from the camera 10 to the background b is longer (further) than the distance from the camera 10 to the face f. In the first image g1 captured with the aperture value of the camera 10 at a predetermined value (for example, F16), the depth of field is relatively deep. The degree of blur (bokeh amount) is relatively small. In the second image g2 captured with the aperture value of the camera 10 smaller than a predetermined value (for example, F2.8), the depth of field is shallow, so that the background b is blurred as shown in FIG. 8B. increase in the degree of

FIG. 9 is a diagram showing an image of a printed facial photograph p captured by the camera 10 in the third embodiment. FIG. 9A shows the first image g1 captured with the aperture value (F-number) of the camera 10 set to a predetermined value (for example, F16), and FIG. For example, the second image g2 captured with F2.8). Note that the camera 10 is focused on the person's face f by autofocus processing. When the subject is a photograph of the face p, the background bp and the face f represented by diagonal lines are on the same plane in the photograph of the face p, and the distance from the camera 10 is the same. Regardless of the size, the blurring degree of the background bp is small.

Therefore, by photographing the same subject while changing the depth of field of the camera 10 and comparing the degree of blurring of the background image around the face area in the photographed image, it is possible to determine spoofing using the photograph of the face or the image of the display. can do.

FIG. 10 is a functional block diagram configured in authentication ECU 200b in the third embodiment. In FIG. 10, the face extraction unit 204, the reference face image acquisition unit 205, and the collation determination unit 206 are the same as those of the authentication ECU 200 of the first embodiment, so description thereof will be omitted.

The camera control unit 201b controls the camera 10 via the camera ECU 100. FIG. For example, when the mobile terminal 500 is held over the reader of the NFC device installed near the door 3 of the vehicle 1, the camera ECU 100 is instructed to capture the first image g1 and the second image g2 by the camera 10. Output. The imaging command is to image the subject with a predetermined aperture value (eg, F16) as the first image g1, and image the subject with an aperture value smaller than the predetermined aperture value (eg, F2.8) as the second image. Directive. The imaging of the first image g1 and the imaging of the second image g2 are continuously performed, and the first image g1 and the second image g2 are images of the same subject.

When the camera ECU 100 receives the imaging command, the focus control unit 101 focuses on the face area of the subject (person). Then, the camera 10 sets the aperture value to a predetermined aperture value, adjusts the shutter speed (exposure time) and the ISO sensitivity so that the exposure control unit 102 obtains an appropriate exposure, and captures the first image g1. to work. The camera 10 outputs the captured image as the first image g1. After capturing the first image g1, the camera ECU 100 continuously sets the aperture value to be smaller than a predetermined aperture value, and the exposure control unit 102 controls the shutter speed (exposure time) and The camera 10 is operated so as to adjust the ISO sensitivity and capture the second image g2. The camera 10 outputs the captured image as the second image g2.

The face/background area detection unit 202b uses a machine learning algorithm to detect the face area of the image from the first image g1 and the second image g2 captured by the camera 10, and specifies the face area. Note that the positions of the eyes, nose, and mouth of the subject may be detected, and based on the position information, the face area of the image may be detected and the face area may be specified. Then, an area within a predetermined range that is at least a predetermined distance away from the identified face area is identified as a background area.

The face/background area detection unit 202b calculates the degree of background image blur (bokeh amount) bk1 from the background area of the first image g1, and calculates the degree of blur of the background image (bk1) from the background area of the second image g2. Bokeh amount) bk2 is calculated. For example, in the background area, the smaller (weaker) the contrast between adjacent pixels, the larger the degree of blurring.

The spoofing determination unit 203b compares the degree of blur bk1 of the background image of the first image g1 and the degree of blur bk2 of the background image of the second image g2 calculated by the face/background area detection unit 202b. To determine whether or not a captured image is a spoofed image using a photograph of the face. When the difference between the degree of blur bk1 and the degree of blur bk2 is equal to or greater than a predetermined value, the image is an image of a person and is determined not to be spoofed, and a signal indicating that the image picked up by the camera 10 is not spoofed is output. If the difference between the degree of blur bk1 and the degree of blur bk2 is smaller than a predetermined value, it is determined that the image is an image obtained by picking up a face photograph and that the image is impersonation. Note that the predetermined value is set in advance by experiment or the like.

FIG. 11 is a schematic flowchart of processing executed by the authentication ECU 200b. In this flowchart, S10 to S13 in the flowchart of FIG. 5 are replaced with S30 to S33. In FIG. 11, the processes of S14 to S17 are the same as those in the flowchart of FIG. 5, so description thereof will be omitted.

First, in S30, the camera 10 captures an image of the subject to acquire the first image g1. The first image g1 is captured by focusing on the face area of the subject (person), setting the aperture value to a predetermined aperture value, and adjusting the shutter speed (exposure time) and ISO sensitivity to achieve proper exposure. This is an image with

Subsequently, in S31, the subject is imaged by the camera 10 and the second image g2 is obtained. For the second image g2, the face area of the subject (person) is focused, the aperture value is set to an aperture value smaller than a predetermined aperture value, and the shutter speed (exposure time) and ISO sensitivity are adjusted to achieve proper exposure. It is an image captured after adjustment.

In S32, the degree of blur (bokeh amount) bk1 of the background image of the first image g1 and the degree of blur (bokeh amount) bk2 of the background image of the second image g2 are calculated. From the first image g1 and the second image g2, a machine learning algorithm is used to detect and identify the face area of the image. Then, an area within a predetermined range that is at least a predetermined distance away from the identified face area is identified as a background area. Then, in the background area of the first image g1, the degree of blur bk1 is calculated from the contrast between adjacent pixels. Similarly, in the background area of the second image g2, the degree of blur bk2 is calculated from the contrast between adjacent pixels.

Subsequently, in S33, it is determined whether or not the difference between the degree of blur bk1 and the degree of blur bk2 is equal to or greater than a predetermined value. If the difference between the degree of blur bk1 and the degree of blur bk2 is greater than or equal to the predetermined value, the determination is affirmative and the process proceeds to S14. If the difference between the degree of blur bk1 and the degree of blur bk2 is smaller than the predetermined value, a negative determination is made and the current routine ends.

According to the third embodiment, the first image g1 is captured with a predetermined aperture value (eg, F16), and the second image g2 is captured with a smaller aperture value than the predetermined aperture value (eg, F2.8). . Then, when the difference between the degree of blur bk1 of the background image of the first image g1 and the degree of blur bk2 of the background image of the second image g2 is equal to or greater than a predetermined value, it is determined that the image is not impersonation, and the photographed image is used to detect the face. doing authentication. Therefore, as in Embodiments 1 and 2, spoofing can be determined and spoofing can be suppressed without providing a dedicated device for capturing the features of a face photograph or a display image.

As described above, according to the above-described embodiments (Embodiment 1, Embodiment 2, and Embodiment 3), for the same subject, the subject is focused, and parameters related to exposure are A first image and a second image are captured in different states. Then, from the first image and the second image, the background, which is the peripheral portion of the face area, is extracted as a background image, and spoofing is determined using the background image of the first image and the background image of the second image. As a result, spoofing can be determined and spoofing can be suppressed without providing a dedicated device for capturing the characteristics of the face photograph or the image on the display.

In the above embodiment, when the mobile terminal is held over the reading unit of the NFC device, the camera captures an image of the subject and acquires the first image and the second image. However, face authentication may be performed without using a mobile terminal. For example, a motion sensor provided near the camera may detect that the member who has made the rental reservation approaches the camera, and the subject (person) may be photographed. This makes it possible to carry out authentication without carrying a mobile terminal.

In the above embodiment, the face/background area detection unit 202 (202a, 202b) detects the face area of the image from the image captured by the camera 10 to specify the face area, and from the image of the specified face area, , the face extraction unit 204 extracts feature amount data Fi (face image) representing features of the face. However, when using a machine learning algorithm capable of specifying a face region and extracting feature amount data Fi (face image) indicating facial features, the function of the face extraction unit 204 is replaced by face/background region detection. It may be integrated into the section 202 (202a, 202b).

In the above embodiment, the reference face image (feature amount data Ri) is acquired from the management server 400. FIG. However, it may be stored in the storage device of the vehicle 1, for example, the memory of the authentication ECU. In this case, the authentication ECU can perform face authentication without communicating with the management server 400 .

In the above embodiment, the facial image (feature data Fi) of the person to be authenticated is extracted from the first image or the second image taken for judging impersonation. Therefore, spoofing determination and face authentication can be performed at the same time. It should be noted that in the above embodiment, the face image of the person to be authenticated is detected from the first image or the second image, and face authentication is performed. However, when the face image of the person to be authenticated is detected from both the first image and the second image, and the face image detected from both images matches the reference face image, face authentication is established. good too.

In the above embodiment, the first image and the second image that are captured in succession are used. However, three or more images may be used for judging impersonation as long as images of the same subject are continuously captured. Alternatively, an image may be captured while continuously adjusting the exposure (exposure value) and the amount of light (brightness of the illumination), and spoofing may be determined from the profile of the difference between the background images.

The embodiments disclosed this time should be considered as examples and not restrictive in all respects. The scope of the present invention is indicated by the scope of the claims rather than the description of the above-described embodiments, and is intended to include all modifications within the scope and meaning equivalent to the scope of the claims.

1 vehicle, 2 vehicle body, 3 door, 5 authentication system, 10 camera, 20 electronic flash, 30 door lock device, 100 camera ECU, 101 focus control unit, 102 exposure control unit, 103 light emission control unit, 200, 200a, 200b authentication ECU, 201, 201b camera control unit, 202, 202a, 202b face/background area detection unit, 203, 203a, 203b spoofing determination unit, 204 face detection unit, 205 reference face image acquisition unit, 206 matching determination unit, 250 communication unit , 300 door ECU, 400 management server, 401 ECU, 402 storage device, 403 communication device, 500 portable terminal.

Claims (10)

A face authentication system that performs authentication by matching a face image captured by a camera with a reference face image,
a camera control unit configured to control the camera to capture a first image and a second image having different exposure-related parameters with respect to the same subject, focusing on the subject;
a background area detection unit that detects, as a background image, a background that is a peripheral portion of a face area in the first image and the second image;
A face authentication system comprising: a spoofing determination unit that determines spoofing using the background image of the first image and the background image of the second image. The camera control unit captures the first image with a predetermined exposure, captures the second image with illumination brighter than the illumination when the first image was captured, and with an exposure value greater than the predetermined exposure,
2. The face authentication system according to claim 1, wherein said spoofing determination unit determines that spoofing is not performed when a difference in brightness between said background image of said first image and said background image of said second image is equal to or greater than a predetermined value. . The camera control unit captures the first image with a predetermined exposure, captures the second image with illumination brighter than the illumination when the first image was captured, and with an exposure value greater than the predetermined exposure,
The spoofing determination unit determines that a difference in brightness between the image of the face area in the first image and the background image is smaller than a predetermined value, and a difference in brightness between the image of the face area in the second image and the background image. 2. The face authentication system according to claim 1, wherein when is equal to or greater than the predetermined value, the face authentication system determines that it is not spoofing. The camera control unit captures the first image with a predetermined aperture value and captures the second image with an aperture value smaller than the predetermined aperture value,
2. The face authentication according to claim 1, wherein said spoofing determination unit determines that spoofing is not performed when a difference in degree of blur between said background image of said first image and said background image of said second image is equal to or greater than a predetermined value. system. a face extraction unit that extracts the face image from the image captured by the camera;
a matching determination unit for matching the face image and the reference face image to perform face authentication;
The collation determination unit compares the face image extracted from at least one of the first image and the second image with the reference face image to perform face authentication when the impersonation determination unit determines that the impersonation is not performed. 5. The face authentication system according to any one of claims 2 to 4, wherein: A face authentication method for performing authentication by matching a face image captured by a camera with a reference face image,
capturing with the camera a first image and a second image with different exposure-related parameters, focusing on the same subject;
a step of detecting a background, which is a peripheral portion of a face area, in the first image and the second image as a background image;
determining spoofing by comparing the background image of the first image and the background image of the second image. The step of capturing images with the camera includes capturing the first image with a predetermined illumination and a predetermined exposure, and capturing the second image with illumination brighter than the predetermined illumination and an exposure value greater than the predetermined exposure. including steps and
7. The face according to claim 6, wherein the step of judging whether or not it is spoofing determines that it is not spoofing when a difference in brightness between said background image of said first image and said background image of said second image is equal to or greater than a predetermined value. Authentication method. The step of capturing images with the camera includes capturing the first image with a predetermined illumination and a predetermined exposure, and capturing the second image with illumination brighter than the predetermined illumination and an exposure value greater than the predetermined exposure. including steps and
The step of judging spoofing comprises: a difference in brightness between the image of the face area of the first image and the background image being smaller than a predetermined value, and the brightness of the image of the face area of the second image and the background image; 7. The face authentication method according to claim 6, wherein when the difference between is equal to or greater than a predetermined value, it is determined that the person is not impersonating. The step of capturing an image with the camera includes capturing the first image with a predetermined aperture value, and capturing the second image with an aperture value smaller than the predetermined aperture value,
7. The step of determining spoofing according to claim 6, wherein when a difference in degree of blur between the background image of the first image and the background image of the second image is equal to or greater than a predetermined value, it is determined that the image is not spoofing. face recognition method. In the step of determining spoofing, when it is determined that spoofing is not performed, the face image extracted from at least one of the first image and the second image is compared with the reference face image to perform face authentication. 10. A face authentication method according to any one of claims 7 to 9, further comprising .

Images