JP2019159372A - Authentication apparatus, authentication system, authentication method, computer program and storage medium - Google Patents

Abstract

To provide an authentication apparatus that performs high-accuracy determination by preventing impersonation due to forgery of physical features while reducing burden of a user.SOLUTION: An authentication apparatus 1000 includes an image acquisition unit 1200 for acquiring a picked-up image picking up a user from an imaging apparatus 300, an image record unit 1300 for recording a register image that is an image of a registered user, a collation area determination unit 1400 for determining a collation area for collating a part of the picked-up image and a part of the registered image, and a collation unit 1500 for collating an image in the collation area of the picked-up image and the image in the collation area of the registered image and determining whether a user in the picked-up image is the registered user.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an authentication technique using an image captured by an imaging device.

  There is biometric authentication that determines a person having a physical feature based on the physical feature of the person. Physical features used for biometric authentication include personal movement features (behavioral features) such as gaits, handwriting, and voiceprints, and personal organs such as faces, irises, retinas, fingerprints, finger veins, and palms. There are shape features (physical features). In biometric authentication, a physical feature registered in advance is compared with a physical feature acquired from a person to be determined, and the degree of similarity is derived. If the similarity is greater than or equal to a predetermined value, it is determined that the person to be determined is a registered person, and if the similarity is less than a predetermined value, it is determined that the person to be determined is not registered.

  A typical embodiment of biometric authentication is entrance / exit management that allows only authorized persons to enter the room. In entrance / exit management, high determination accuracy is required such that a person who has the authority to enter the room is not rejected by mistake, and a person who does not have the authority to enter the room is not erroneously permitted. As biometric authentication technologies that achieve high determination accuracy, iris authentication technology (Patent Literature 1), fingerprint authentication technology (Patent Literature 2), finger vein authentication technology (Patent Literature 3), and the like have been widely put into practical use.

  In entrance / exit management, a person who wants to enter a room causes a sensor installed near the entrance of the room to enter the room to detect his / her own physical features (eyes, fingers, etc.). The authentication device collates the physical characteristics detected by the sensor with the physical characteristics registered in advance, and determines whether or not to allow the person who wants to enter the room to enter the room. In such a conventional configuration, a predetermined action is forced on the person who wants to enter the room so that the sensor can always detect the physical characteristics under a certain condition. Thereby, high determination accuracy is realized.

  In biometric authentication, measures against counterfeiting of physical features are important. If a third party forges a physical feature, the third party may be authenticated as a registered person. In the research fields of fingerprint authentication and iris authentication, studies have been reported in the case where physical features are forged (Non-Patent Documents 1, 2, and 3).

JP 2002-259981 A JP 2007-117535 A JP 2007-156936 A

“Does the fingerprint verification device accept artificial fingerprints?”, Koji Yamada, Hiroyuki Matsumoto, Tsutomu Matsumoto, IEICE Technical Report. ISEC, Information Security vol.100, no.213, pp.159-166, 2000-07 -18 "What do you do when you find weaknesses in security technology?" Tsutomu Matsumoto, IEICE Journal 00084 (00003), 202-204, 2001-03-01 "From the iriscode to the iris: a new vulnerability of irisrecognition systems": J. Galbally: 2012

  Biometric authentication is desired to reduce the load on the user while maintaining high determination accuracy. For example, in the entrance / exit management using the iris, fingerprint, and finger vein authentication as described above, the user needs to cause the sensor to detect a physical feature. Specifically, the user needs to approach the sensor near the entrance and move the physical feature within the detection range of the sensor. The user may spend tens of seconds on the operation for cooperating with such authentication. In particular, in fingerprint and finger vein authentication, when a user is wearing gloves, it must be removed, which is a burden on the user.

  Also, authentication using a face, iris, retina, fingerprint, finger vein, palm shape, etc. uses a specific body part as a physical feature. For this reason, these physical characteristics are fundamentally irreplaceable. Disabling a forged physical feature as a countermeasure against forgery of a physical feature can prevent impersonation of a third party, but also prevents the registered original user himself from being authenticated. Patent Document 3 proposes a method of suppressing authentication breakthrough due to counterfeiting by randomly using one of a plurality of fingerprint images taken from a plurality of directions for authentication. However, because the body parts used for authentication are the same, it is not an essential solution to counterfeiting physical features.

  SUMMARY OF THE INVENTION In order to solve such a conventional problem, the present invention mainly provides an authentication device that performs high-accuracy determination by preventing impersonation due to forgery of physical features while reducing the load on the user. And

  An authentication apparatus according to the present invention includes an image acquisition unit that acquires a captured image obtained by capturing a person, a recording unit that records a registered image that is an image of a person to be registered, a part of the captured image, and one of the registered images. A collation area determination means for determining a collation area for collating the image and a collation area image of the captured image and a collation area image of the registered image, and a person of the captured image is registered Collating means for judging whether or not the person is a person.

  ADVANTAGE OF THE INVENTION According to this invention, imitation by the forgery of a physical feature can be prevented and a highly accurate determination can be performed, reducing a user's load.

(A), (b) is a block diagram of an authentication system. (A), (b) is explanatory drawing of the installation position of an authentication system. The flowchart showing an authentication process. The flowchart showing collation area | region determination processing. (A)-(c) is explanatory drawing of a collation area | region determination process. (A), (b) is explanatory drawing of the installation position of an authentication system.

  Hereinafter, embodiments will be described in detail with reference to the drawings.

(First embodiment)
In this embodiment, an example in which the authentication device is used for entrance / exit management will be described. Here, an image obtained by capturing a person's profile is used for the authentication process. Hereinafter, a person who wants to enter the room is simply referred to as a “user”, and a person who is given authority to enter and leave the room in advance is referred to as a “registered user”.

(overall structure)
FIG. 1 is a configuration explanatory diagram of an authentication system including an authentication apparatus according to the present embodiment. FIG. 1A is a hardware configuration diagram of the authentication system 1. The authentication system 1 includes an authentication device 1000, an approach detection sensor 200, an imaging device 300, an input device 400, and an unlocking device 500. The proximity detection sensor 200, the imaging device 300, the input device 400, and the unlocking device 500 are connected to the authentication device 1000 by wire or wirelessly.

The authentication apparatus 1000 includes a CPU (Central Processing Unit) 110 and a ROM (Read
The computer system includes only memory (RAM) 120 and random access memory (RAM) 130. The authentication device 1000 includes a storage device 140, an input / output interface (I / F) 150, an input I / F 160, and an output I / F 170. The CPU 110, the ROM 120, the RAM 130, the storage device 140, the input / output interface (I / F) 150, the input I / F 160, and the output I / F 170 are connected to be communicable with each other via a bus.

  The CPU 110 controls the operation of the authentication system 1 by executing a computer program stored in the ROM 120 using the RAM 130 as a work area. The storage device 140 is a mass storage device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive). The storage device 140 stores physical features such as facial images of registered users who are permitted to enter a room that manages entry / exit. The input / output I / F 150 performs interface control between the authentication device 1000, the proximity detection sensor 200, and the imaging device 300. The input / output I / F 150 transmits a control signal for controlling operations to the proximity detection sensor 200 and the imaging apparatus 300 under the control of the CPU 110. The input / output I / F 150 acquires the detection result of the proximity detection sensor 200 and the image captured by the imaging device 300 and transmits the acquired image to the CPU 110.

  The approach detection sensor 200 is a sensor that is installed in the vicinity of an entrance of a room that manages entry / exit, and detects a person (user). The proximity detection sensor 200 is a human sensor such as an infrared sensor. The imaging apparatus 300 is installed near the entrance of the room that manages the entrance and exit of the room and closer to the entrance than the proximity detection sensor 200. The imaging device 300 images a user approaching the room.

  The input I / F 160 performs interface control between the authentication device 1000 and the input device 400. The input device 400 is a user interface such as a keyboard, a mouse, and a touch panel. The input device 400 is used for inputting personal information of the user when the user is registered. The input I / F 160 transmits data input from the input device 400 to the CPU 110. The output I / F 170 performs interface control between the authentication device 1000 and the unlocking device 500. The output I / F 170 transmits an unlocking control signal for controlling the operation of the unlocking device 500 under the control of the CPU 110. The unlocking device 500 unlocks the entrance of a room that manages entry / exit. The entrance of the room that manages entry / exit is always locked. When the unlocking device 500 acquires the unlocking control signal from the authentication device 1000, the unlocking device 500 unlocks the entrance.

  FIG. 1B is a functional block diagram of the authentication apparatus 1000. The authentication device 1000 functions as an approach detection unit 1100, an image acquisition unit 1200, an image recording unit 1300, a collation area determination unit 1400, a collation unit 1500, an external output unit 1600, and an image registration unit 1700. Each functional block is realized by, for example, the CPU 110 executing a predetermined computer program, but at least a part may be realized by hardware.

  The approach detection unit 1100 is realized by the input / output I / F 150 and the CPU 110, and detects a user near the entrance of the room according to a detection result acquired from the approach detection sensor 200. The image acquisition unit 1200 is realized by the input / output I / F 150 and the CPU 110, and acquires a captured image from the imaging apparatus 300. The image acquisition unit 1200 acquires a captured image from the imaging device 300 when the approach detection unit 1100 detects a user near the entrance of the room. The imaging apparatus 300 may be configured to always capture images, but in the present embodiment, the imaging apparatus 300 is configured to capture a high-quality image when there is a request for a captured image from the image acquisition unit 1200. The image recording unit 1300 is realized by the storage device 140 and records registration data including physical characteristics of a registered user registered in the authentication system 1. The image recording unit 1300 of this embodiment records an image (registered image) of a registered user who has the authority to enter and leave the room in association with personal information such as the name of the registered user.

  The collation area determination unit 1400 is realized by the CPU 110, and determines a partial area (collation area) of the captured image and the registered image that are actually collated at the time of collation processing for user authentication. The collation unit 1500 is realized by the CPU 110 and collates the respective collation areas of the captured image and the registered image to determine whether or not the user is a registered user. The external output unit 1600 is realized by the output I / F 170 and the CPU 110, and transmits an unlock control signal to the unlocking device 500 when the user is a registered user as a result of collation. The image registration unit 1700 is realized by the input I / F 160 and the CPU 110, acquires physical characteristics (registered images), personal information, and the like of the user to be registered from the input device 400 and registers them in the image recording unit 1300.

  FIG. 2 is an explanatory diagram of an installation position of the authentication system 1. FIG. 2A is a diagram of the user viewed from the horizontal direction, and FIG. 2B is a bird's-eye view of the user. In the present embodiment, the proximity detection sensor 200 and the imaging device 300 are installed in a passage that leads to an entrance 2001 of a room that is an entry / exit management target. The proximity detection sensor 200 is provided at a position farther from the entrance 2001 than the imaging device 300. Therefore, when the user approaches the entrance 2001, the user passes through the imaging range 2006 of the imaging device 300 after passing through the detection position 2005 of the proximity detection sensor 200. The proximity detection sensor 200 and the imaging device 300 are arranged apart from each other by a distance 2002 corresponding to the time from when the proximity detection sensor 200 detects a user until the imaging device 300 starts imaging.

  The authentication system 1 performs authentication processing so that the lock of the entrance 2001 is unlocked before the timing when the user passes the entrance 2001. Therefore, the imaging device 300 is provided at a position away from the entrance 2001 by a distance 2003 corresponding to the time required from the authentication process to unlocking. The imaging device 300 is arranged at a height that matches the height of a typical user's face in order to capture the user's face. Note that a plurality of imaging devices 300 may be installed along the direction in which the user moves. When a plurality of imaging devices 300 are installed, the imaging device closest to the entrance 2001 is provided at a position away from the entrance 2001 by a distance 2003. By installing a plurality of imaging devices 300, it is possible to capture a plurality of times in the process of passing by the user, so that the authentication accuracy can be increased. The width 2004 of the passage may be a distance that allows the imaging apparatus 300 to capture the feature of the user's skin.

(Authentication process)
FIG. 3 is a flowchart showing an authentication process by the authentication system 1 having the above configuration. Here, in order to simplify the description, it is assumed that only one user is captured in the captured image captured by the imaging apparatus 300. In actual operation, a plurality of users may be captured in the captured image. In this case, this flowchart is repeatedly executed for the number of users captured in the captured image.

  The approach detection unit 1100 of the authentication device 1000 detects that the user has passed through the passage to the entrance 2001 of the room to be entered / exited according to the detection result of the approach detection sensor 200 (S1101). The authentication device 1000 starts the authentication process using a user detection by the proximity detection sensor 200 as a trigger. In response to the proximity detection unit 1100 detecting the user, the image acquisition unit 1200 causes the imaging device 300 to capture an image (S1102).

  The collation area determination unit 1400 acquires one registered image from the image recording unit 1300 (S1103). The collation area determination unit 1400 determines a partial area (collation area) to be collated from each of the captured image and the registered image (S1104). The collation unit 1500 extracts and collates the image of the determined collation area from each of the captured image and the registered image (S1105). The authentication apparatus 1000 performs collation processing of all registered images and captured images recorded in the image recording unit 1300 by repeating the processing of S1103 to S1105 (S1106: N).

  When the collation process between all registered images recorded in the image recording unit 1300 and the captured image is completed (S1106: Y), the collation unit 1500 determines that the user of the captured image is based on the collation result between the captured image and each registered image. It is finally determined whether or not the user is a registered user (S1107). If the result of determination by the collation unit 1500 is that the user of the captured image is a registered user, the external output unit 1600 transmits an unlock control signal to the unlocking device 500 to unlock the entrance 2001. When the user of the captured image is not a registered user, the authentication apparatus 1000 ends the process without transmitting the unlock control signal. For this purpose, the inlet 2001 remains locked. When a display is connected to the authentication apparatus 1000, the external output unit 1600 displays the result of determination by the collation unit 1500 on the display (S1108).

  In the above processing, the final determination processing (S1107) is performed after the collation processing for all registered images and captured images is completed. However, the present invention is not limited to this, and the collation between registered images and captured images is performed. The determination process may be performed every time the process ends. That is, the order of the process of S1106 and the process of S1107 may be switched. In this case, when the user of the captured image is a registered user, the entrance 2001 is unlocked at that time. Therefore, the authentication process is performed more efficiently than when all registered images and captured images are collated.

(Details of each functional block)
The authentication apparatus 1000 authenticates the user as described above, and performs unlocking control of the entrance 2001 according to the authentication result. The details of each functional block and each process of the authentication apparatus 1000 will be described below.

-Approach detection unit 1100
As described above, the approach detection unit 1100 detects a user approaching the entrance 2001 according to the detection result of the approach detection sensor 200. When the moving user crosses the detection position 2005, the approach detection sensor 200 outputs a detection signal indicating that the user has been detected. The approach detection unit 1100 acquires this detection signal and detects that the user has reached the detection position 2005. When the approach detection unit 1100 detects that the user has reached the detection position 2005, the approach detection unit 1100 notifies the image acquisition unit 1200 of the detection of the user. The approach detection sensor 200 can be realized by, for example, an optical sensor used in a railroad crossing obstacle detection device operated on a railway, or a general camera.

  When the optical sensor is used, a light emitting device that emits infrared light is provided on one wall surface of the passage, and a light receiving device that receives infrared light emitted from the light emitting device is provided on a wall surface facing the passage. An infrared optical path from the light emitting device to the light receiving device is a detection position 2005. Since the infrared light is blocked by the user passing through the detection position 2005, the light receiving device cannot receive the infrared light. The approach detection unit 1100 acquires detection signals having different values depending on whether the light receiving device receives infrared rays or not. The approach detection unit 1100 can detect a user based on a detection signal when the light receiving device does not receive infrared rays.

  When a general camera is used, the approach detection unit 1100 detects a user by analyzing an image captured by the camera. The camera takes a moving image, for example. The approach detection unit 1100 acquires a moving image as a detection signal from the camera. The approach detection unit 1100 can detect a user by detecting a face from a moving image. A known technique can be used for face detection. A known technique is, for example, “Rapid object detection using a boosted cascade of simple features: P. Viola, M. Jones: 2001”.

Image acquisition unit 1200
The image acquisition unit 1200 acquires, from the imaging device 300, a captured image obtained by capturing the user moving in the passage, from the acquisition of the notification indicating that the user has been detected from the approach detection unit 1100. The image acquisition unit 1200 transmits the acquired captured image to the collation region determination unit 1400. The imaging device 300 includes an optical lens and an image sensor. The image acquisition unit 1200 controls the operation of the imaging device 300. The image acquisition unit 1200 waits for a predetermined time after acquiring the notification from the proximity detection unit 1100 according to the installation interval (distance 2002) between the imaging device 300 and the proximity detection sensor 200 and the typical walking speed of the user. To do. After waiting, the image acquisition unit 1200 instructs the imaging device 300 to acquire an image and acquires a captured image.

  In addition to the configuration in which the imaging device 300 is provided on one wall surface of the passage and images the user from the side, the imaging device 300 may be provided on each of the opposing wall surfaces of the passage and configured to image the user from both directions. The imaging apparatus 300 may capture the user at an oblique angle of view with a slight angle rather than directing the angle of view in the horizontal direction. As a result, the imaging apparatus 300 can also capture a shadowed person when a plurality of persons pass through the imaging range 2006 simultaneously.

  The imaging apparatus 300 may be a camera that captures a visible image using general visible light, but may be a camera that captures an image using near infrared rays. The imaging device 300 using the near-infrared wavelength can capture a feature of the skin, not the surface of the skin. In addition, there is a report that in the narrow wavelength band between red and near-infrared rays, the reflectance of the skin is significantly different depending on the presence or absence of melanin. This is reported, for example, in “The optics of human skin Aspects important for human health: K. P. Nielsen etc .: 2008”. The imaging device 300 may be a camera that performs imaging in this wavelength band.

  The imaging device 300 may be used in combination with a projector. The projector irradiates the imaging range 2006 with light. Thereby, the imaging apparatus 300 can perform imaging with abundant light quantity. Therefore, the imaging apparatus 300 can capture an image with less movement blur of the user by capturing with an increased shutter speed. In addition, the imaging apparatus 300 can capture a high-quality image with less noise by capturing with low ISO sensitivity. For example, the projector is preferably disposed adjacent to the imaging device 300 and irradiates the user with light from the side at the imaging timing of the imaging device 300. By irradiating the user with light from the side, it is possible to prevent the light from the projector from directly entering the eyes of the user. The spectral characteristics of the projector are preferably light close to the wavelength band received by the image sensor of the imaging apparatus 300. The projector emits visible light when the imaging apparatus 300 captures a visible image, and emits infrared light when capturing an infrared image.

Collation area determination unit 1400
The collation area determination unit 1400 determines the collation area used for the collation process as described above. The collation area determination process in S1104 will be described in detail. The matching area determination unit 1400 determines a matching area for the captured image acquired from the image acquisition unit 1200 and the registered image acquired from the image recording unit 1300. FIG. 4 is a flowchart showing the collation area determination process. FIG. 5 is an explanatory diagram of the collation area determination process.

  The collation area determination unit 1400 detects an area where a face exists in the image from the captured image and the registered image (S1201). Face detection can be performed by the above-described known technique. The collation area determination unit 1400 detects a face area that is a rectangular area including the area where the face exists by face detection. The collation area determination unit 1400 may detect a position where a facial organ such as an eye, a mouth, or an ear is present, and specify the position of the face in more detail. By detecting the position of the facial organ, the subsequent processing can be performed with high accuracy. The position of the facial organ can be detected by a known method. For example, “Active Shape Models-Their Training and Application: T.F. Cootes, C.J. Taylor, D. Cooper, and J. Graham: 1998”.

  The collation area determination unit 1400 normalizes the captured image and the registered image by converting the coordinates of the face area to predetermined coordinates (S1202). The collation area determination unit 1400 performs coordinate conversion by performing affine transformation of the face area image so that the face area has a rectangular horizontal width and vertical width of a predetermined length and the face area is erect, for example. When the positions of facial organs such as eyes, mouth, and ears are detected, the matching region determination unit 1400 performs affine transformation so that the positions of the respective organs move to predetermined positions. By the coordinate conversion, the image of the face area becomes a normalized image as illustrated in FIG.

  The collation area determination unit 1400 detects an area where the skin is exposed (skin area) from the normalized face area image of the captured image and the registered image (S1203). The skin region is detected by a known method based on the color in the image. For example, “Statistical color models with application to skin detection: M.J. Jones, J.M. Rehg: 2002”. When the skin area cannot be acquired (S1204: N), the collation area determination unit 1400 determines that the authentication has failed and ends the process of S1104. This is because there is a high possibility that the user has passed through the imaging range 2006 without exposing the skin with an attachment or the like. In this case, the authentication process itself ends, and the entrance 2001 is not unlocked.

  When the skin area is acquired (S1204: Y), the matching area determination unit 1400 extracts candidate areas that are candidates for the matching area from the normalized face area images of the captured image and the registered image (S1205). . Extraction of candidate areas is performed as follows, for example. As illustrated in FIG. 5B, the matching region determination unit 1400 divides the normalized face region image into grids, and extracts only the skin region grid. In FIG. 5B, a hatched grid is extracted. Each grid of the extracted skin area becomes a candidate area. The size of the grid is, for example, a rectangle with an actual size and a side of 2 to 3 centimeters. When the size of the grid is too small, the amount of information for identifying an individual is insufficient. When the size of the grid is too large, it is difficult to extract a grid that is a skin region due to the influence of the face outline and hair. Therefore, the grid size is set appropriately.

  The collation area determination unit 1400 extracts candidate areas having overlapping positions from candidate areas extracted based on the normalized face area images of the captured image and the registered image (S1206). Candidate areas with overlapping positions are candidate areas that can be collated at the same position in each image. FIG. 5C illustrates candidate regions whose positions overlap. The hatched grid is a candidate area with overlapping positions. Here, a candidate area from the temple to the cheek is extracted as a candidate area whose position overlaps.

  The collation area determination unit 1400 selects one predetermined candidate area from candidate areas whose positions overlap (S1207). For example, the matching area determination unit 1400 may select candidate areas at different positions according to registered users. Or the collation area | region determination part 1400 selects the candidate area | region of a different position at random, whenever a captured image is acquired. In any case, the matching area determination unit 1400 does not fix the position of the candidate area to be selected, and selects candidate areas at various positions from candidate areas where a plurality of positions overlap.

  The collation area determination unit 1400 performs fine adjustment for aligning the position corresponding to the selected candidate area in each of the captured image and the registered image (S1208). In the processing of S1201 to S1207, the candidate area can be expected to have the profile in the captured image and the registered image at the same position by face detection. However, there may be misalignment in millimeters. For accurate matching, color misregistration in millimeters also needs to be adjusted.

  In order to collate exactly the same position between the captured image and the registered image, the collation area determination unit 1400 uses the appearance characteristics such as skin surface spots, pores, freckles, moles, etc. Fine-tune the size. The fine adjustment of the alignment is performed by the following simplest method, for example. The matching area determination unit 1400 extracts partial images by slightly shifting the captured image and registered image candidate areas, normalizes the partial images, and performs cross-correlation to perform alignment. From the obtained correlation value, the collation area determination unit 1400 can estimate how much the candidate areas of the captured image and the registered image match. If the position of the candidate region of one image is examined while being shifted little by little, the most suitable position can be searched. When the correlation value finally exceeds the predetermined value, the collation region determination unit 1400 determines that the alignment of the candidate region has been completed. It is desirable to adjust the predetermined value in advance before operation.

  The collation area determination unit 1400 performs fine adjustment of the alignment until at least one candidate area in which the position is matched between the captured image and the registered image is obtained. That is, the collation area determination unit 1400 repeatedly executes the processes of S1207 and S1208 until a candidate area with a matching position is obtained from candidate areas with overlapping positions. When a candidate area that matches the position is obtained (S1209: Y), the collation area determination unit 1400 determines that the alignment is successful. The matching area determination unit 1400 sets a candidate area that has been successfully aligned as a matching area. The matching area determination unit 1400 transmits the position of the matching area corresponding to each of the captured image and the registered image to the matching unit 1500.

  If a candidate area that matches the position cannot be obtained (S1209: N), the collation area determination unit 1400 determines that the alignment has failed. The collation area determination unit 1400 may continue the process using the candidate area that is the best positioned as the collation area, but may not collate the captured image with the registered image. The reason for not performing the verification is that the positions of the captured image and the registered image may be different from each other because the position could not be adjusted based on the appearance of the skin surface such as spots, pores, freckles, and moles. This is because the nature is high. In this case, the authentication process itself ends, and the entrance 2001 is not unlocked.

  In the above description, the collation area determination unit 1400 performs the collation area determination process every time the image acquisition unit 1200 acquires a captured image. On the other hand, the collation area determination unit 1400 may hold the information of the registered image in the image recording unit 1300 in a state in which the process up to S1205 for extracting the candidate area from the registered image is executed in advance. Thereby, the collation area | region determination part 1400 does not need to repeat the same process, and can speed up collation area | region decision processing. Note that the matching area determination unit 1400 may determine a plurality of matching areas. By using a plurality of collation areas, the reliability of collation is improved.

・ Verification unit 1500
The collation unit 1500 performs collation processing and determination processing between the captured image and the registered image in steps S1105 to S1107. The collation unit 1500 first extracts a feature amount used for collation from the captured image acquired from the image acquisition unit 1200 and the registered image acquired from the image recording unit 1300. The collation unit 1500 collates the feature quantities to derive a similarity indicating the same person-likeness between the registered user and the user shown in the captured image. The collation unit 1500 derives the similarity of the registered images of all registered users, and determines whether or not the user shown in the captured image is a registered user. Hereinafter, this will be described in detail.

  The feature amount is extracted as follows. The matching unit 1500 extracts feature amounts used for matching from the matching regions of the captured image and the registered image. Here, the feature quantity is a multidimensional quantity and is represented as a feature vector. The feature amount is, for example, a numerical sequence (vector) in which pixel values in a rectangular matching area are sequentially read out from the upper left of the rectangle toward the lower right. Further, the method of “Face recognition with local binary patterns: T Ahonen, A Hadid, M Pietik ä inen: 2004” may be used before reading out the pixel values. By converting the numerical sequence into only the magnitude relationship between the pixel values of adjacent pixels, it is possible to remove fluctuations in the brightness value of the entire image that change in the imaging environment. Furthermore, various image processing filters can be combined as preprocessing for feature amount extraction.

  In the case where the numerical value sequence in which the pixel values are arranged as they are is used as the feature amount, the entire skin region becomes the feature amount, so that the feature amount becomes redundant. When humans actually compare the skin by looking at the skin, local patterns such as skin moles, spots, pores, blood vessels, vellus hair, and scars are important clues for collation. Specifically, it is important how large and how many patterns of density exist on the skin. For example, only the information can be extracted using the technique of “Facial feature point extraction method based on combination of shape extraction and pattern matching: K. Fukui, O. Yamaguchi: 1998”. This technique discloses a filter operation for determining where a pattern having a specific size and shape exists in a captured image.

  For example, assuming that the mole of the face is substantially circular, the collation unit 1500 can obtain the positions of moles of various sizes by preparing filters corresponding to various sizes of circles. This method can also obtain the accuracy of existence corresponding to the density of the mole. Therefore, the collation unit 1500 can set a feature amount that can identify an individual by using as a feature amount the size and density of the mole on the cheek. By arbitrarily changing the shape of the filter, the matching unit 1500 can extract a pattern having another shape in the same manner.

  Derivation of similarity is performed as follows. The collation unit 1500 collates the feature amount of the captured image with the feature amount of the registered image, and calculates the similarity. The similarity degree of this embodiment represents that the feature amount is similar as it is large. That is, the greater the similarity, the more the user of the captured image is the same as the registered user. For example, the matching unit 1500 calculates the cosine of the angle formed by the vectors of the feature amount of the captured image and the feature amount of the registered image, the reciprocal of the Euclidean distance between the feature amount vectors, and the like as the similarity.

  The determination is performed as follows. The collation unit 1500 acquires the similarity of all registered images with respect to the captured image at the time of this processing. Here, collation unit 1500 determines whether or not the user shown in the captured image is a registered user.

  The collation unit 1500 extracts the maximum similarity among the acquired similarities. The collation unit 1500 determines whether the maximum similarity is greater than a predetermined threshold. When the maximum similarity is larger than the threshold, the collation unit 1500 determines that the user shown in the captured image is a registered user of the registered image that has the maximum similarity. When the maximum similarity is smaller than the threshold value, the collation unit 1500 determines that the user shown in the captured image is not a registered user but an unregistered person.

  A threshold value to be compared with the similarity is set in advance. Authentication errors include unauthenticated that erroneously determines a registered user as an unregistered user and misauthenticated that erroneously determines an unregistered user as a registered user. Unauthenticated and mis-authenticated are in a trade-off relationship, and the type of authentication error to be reduced is selected according to the threshold setting status. When the threshold is low, it is easy to determine different persons as the same person, so the possibility of unauthenticated decreases and the possibility of erroneous authentication increases. When the threshold is high, it is difficult to determine different persons as the same person, so that the possibility of unauthenticated becomes high and the possibility of erroneous authentication is reduced.

Image recording unit 1300
The image recording unit 1300 records the physical characteristics of a registered user who has authority to enter and leave, in this case, the profile image of the user and the positional information of a plurality of candidate areas in association with each other. As described above, the image recording unit 1300 is realized by the readable / writable storage device 140. The image recording unit 1300 is preferably implemented by being distributed among a plurality of physically different storage devices 140. The plurality of storage devices 140 are provided in different places in the real world. Further, it is more preferable that a route for transmitting the image and the position information from the collation area determination unit 1400 to the image recording unit 1300 is provided separately. The physical feature of the present embodiment is a set of an image of the user's skin and the position information of the candidate region, and only one of them does not hold as a physical feature. Such an image recording unit 1300 can reduce the possibility that a third party can steal both physical characteristics and position information of candidate areas.

External output unit 1600
As described above, if the user of the captured image is a registered user, the external output unit 1600 transmits an unlock control signal to the unlocking device 500 as a result of the determination by the matching unit 1500. The unlocking device 500 receives the unlock control signal and unlocks the entrance 2001. The external output unit 1600 does not transmit the unlock control signal when the user of the captured image is not a registered user as a result of the determination by the collation unit 1500. For this purpose, the inlet 2001 remains locked.

  When the user of the captured image is not a registered user, the external output unit 1600 may notify the surroundings that a user who is not registered by causing the speaker to sound is approaching the locked entrance 2001. Further, the external output unit 1600 may display the result of determination by the collation unit 1500 on the display. For example, the external output unit 1600 displays a captured image on a display and displays a determination result superimposed on the captured image.

Image registration unit 1700
The image registration unit 1700 registers a user in the image recording unit 1300 based on an input from the input device 400. At the time of user registration, the authentication apparatus 1000 operates in a registration mode. In the registration mode, the authentication device 1000 displays a user registration screen on the display, for example, and guides the registration operation.

The image of the user to be registered is acquired by the image acquisition unit 1200 when the user passes through the imaging range 2006 of the imaging apparatus 300 in the registration mode as in the authentication process. The image registration unit 1700 acquires the registration image acquired in this way from the image acquisition unit 1200.
At this time, the collation region determination unit 1400 executes the processing of S1201 to S1205 for the registration image. Thereby, the collation area | region determination part 1400 extracts a some candidate area | region from the image for registration. The image registration unit 1700 acquires position information of a plurality of candidate areas from the collation area determination unit 1400. The image registration unit 1700 acquires personal information such as user identification information and names to be registered from the input device 400, and stores them in the image acquisition unit 1200 together with position information of a plurality of candidate areas and registration images.

  Since the registered user who has been registered once may have skin changes over time, it is preferable that the physical features (face images) are regularly updated. Therefore, when it is determined that the user is a registered user, the image registration unit 1700 records the captured image at that time in the image recording unit 1300 as a new physical feature (face image). For example, the image registration unit 1700 updates the physical characteristics when a predetermined period has elapsed since the previous registration in the process of S1108 in FIG.

  In this case, the image registration unit 1700 sets a threshold that is higher than the threshold that is compared with the similarity during the determination process, and updates the physical feature only when the similarity exceeds the high threshold. It may be. That is, the image registration unit 1700 updates the physical feature only when it can be authenticated more reliably in order to prevent the physical feature from being mistakenly replaced by another person when updating the physical feature.

  The authentication system 1 according to the present embodiment as described above can realize highly accurate authentication processing without making the user aware of it. The imaging apparatus 300 according to the present embodiment captures a high-resolution image at a position close to the skin of a specific part of the user. In order to perform authentication without making the user aware of it, the authentication apparatus 1000 selects a part that can be imaged at a high resolution according to the conditions of the site to be authenticated, taking advantage of the features of the present invention in which the part used for authentication is not limited. Specifically, in the present embodiment, the imaging device 300 is provided on the wall of the passage to the entrance 2001 that manages entry / exit. The imaging device 300 images the user's face from the side at the moment when the user passes. The authentication device 1000 performs authentication processing on the user's cheek portion of the captured image.

  By arranging the imaging device 300 in a wall rather than in front of the user, the authentication device 1000 can authenticate the user at the timing when the user approaches the imaging device 300 without noticing the imaging device 300. The authentication system 1 according to the present embodiment can perform an authentication process by photographing a hand instead of a user's profile. However, the skin cannot be taken when the user is wearing gloves. If it is a profile, it is difficult to completely cover even if a mask is worn, so there is a high possibility of photographing the skin.

  Moreover, the authentication system 1 of this embodiment makes it difficult to forge physical features. Unlike the biometric authentication such as iris, fingerprint, and finger vein, the authentication system 1 according to the present embodiment uses a captured image captured without contact for authentication. In biometric authentication such as iris, fingerprint, and finger vein, there is a risk that a third party later steals an image or a pattern of the specific part at the timing when the user holds the specific part over the sensor. However, the authentication system 1 of this embodiment does not force the user to perform an authentication operation. Therefore, there is little risk of theft of physical features.

  Furthermore, the authentication system 1 according to the present embodiment can vary the position of the region actually used for authentication among the specific parts while concealing the outside by utilizing the feature that does not limit the parts to be collated. For this purpose, for example, a third party needs to steal and forge a wide area of the cheek. The larger the body part that is stolen or counterfeited, the more difficult it is to forge physical features, as it is generally considered technically difficult.

(Second Embodiment)
2nd Embodiment is an application example of the authentication system 1, and demonstrates the case where the entrance / exit management by the authentication system 1 is not performed explicitly. In the present embodiment, in a facility having an entrance and an exit through which an unspecified number of persons can freely pass, the authentication system 1 measures a staying time from when each person enters the facility until the person leaves. The description of the same configuration and function as the authentication system 1 of the first embodiment is omitted.

  In the second embodiment, a set of the proximity detection sensor 200 and the imaging device 300 is arranged at an entrance and an exit of a facility, respectively. Each approach detection sensor 200 and each imaging device 30 are connected to an authentication device 1000 installed at a physically separated position via a network such as a LAN (Local Area Network).

  The image registration unit 1700 included in the authentication device 1000 does not perform user registration using the input device 400, and automatically registers the captured image captured by the imaging device 300 provided at the entrance as a registered image in the image recording unit 1300. . The authentication device 1000 performs a matching process between a captured image captured by the imaging device 300 provided at the exit and a registered image that is automatically registered.

  The image registration unit 1700 will be described. The image registration unit 1700 registers the captured image captured by the imaging device 300 provided at the entrance as a registered image representing the physical characteristics of a new registered user in the image recording unit 1300. At this time, the image registration unit 1700 records the registered time in the image recording unit 1300 together with the registered image. The authentication device 1000 can calculate the staying time of the facility of the user from the registration time of such a registered image and the time when the user is authenticated using the image captured by the imaging device 300 provided at the exit. it can.

  In the present embodiment, in order to automatically register an unspecified number of persons, the number of registered users may increase without limit. For this purpose, it is preferable that the image registration unit 1700 automatically deletes the registered image and the registration time after a predetermined time has elapsed after recording the image in the image recording unit 1300. The predetermined time is set based on a typical time during which the user stays at the facility.

  Facilities usually do not distinguish between an exit and an entrance. The user is not determined in which direction to move through the passage. Even in such a case, the authentication system 1 of the second embodiment can be applied. Specifically, the authentication device 1000 automatically registers images captured by the imaging devices 300 provided at all entrances and exits of the facility and performs authentication processing using the captured images. However, in order to prevent double registration, the authentication apparatus 1000 performs collation processing with a registered image that has already been registered before automatic registration of the captured image. The authentication apparatus 1000 registers the captured image after determining that there is no corresponding registered user.

  The authentication system 1 according to the second embodiment as described above is a measurement system for a user's facility stay time to which the authentication system 1 according to the first embodiment is applied. The authentication system 1 can measure the user's stay time without being noticed by the user.

(Third embodiment)
The third embodiment is an application example of the authentication system 1, and even if there is a large amount of traffic and a plurality of users overlap in the imaging range 2006 of the imaging device 300 and pass simultaneously, authentication that can authenticate all users is possible. The system 1 will be described. The description of the same configuration and function as the authentication system 1 of the first embodiment is omitted.

  In the third embodiment, unlike the first embodiment, the entrance 2001 is not locked, and a simple gate through which the user can pass without stopping is provided in the passage. The user passes through the entrance 2001 without stopping. The authentication system 1 authenticates whether the user is authorized to pass through the entrance 2001 before passing through the gate.

  FIG. 6 is an explanatory diagram of an installation position of the authentication system 1 according to the third embodiment. Fig.6 (a) is the figure which looked at the user from the horizontal direction, and FIG.6 (b) is a bird's-eye view of a user. The authentication system 1 of this embodiment authenticates a user through a passage with a large traffic volume. For this purpose, the gate is provided with a plurality of gates 2000 that prompt the user to move in a line. The interval between the gates 2000 is preferably a width that allows one user to pass. In this case, the user moves in a line. Note that the shape of the gate 2000 is not limited as long as the user can be prompted to move in a line. For example, the gate 2000 may have a shape such as an upright pole.

  The proximity detection sensor 200 and the imaging device 300 are provided in each of the plurality of gates 2000. The proximity detection sensor 200 is disposed on the upstream side of the imaging device 300 with respect to the moving direction of the user. Therefore, when the user passes through the gate 2000, the user passes through the imaging range 2006 of the imaging device 300 after passing through the detection position 2005 of the proximity detection sensor 200. The proximity detection sensor 200 and the imaging device 300 are arranged separated by a distance 600. The distance 600 is about several tens of centimeters, for example. This distance 600 is a distance according to the time from when the proximity detection sensor 200 detects the user until the imaging device 300 starts imaging. The imaging device 300 is arranged at a distance 601 from the end of the gate 2000. The distance 601 is a distance necessary to authenticate the user before passing through the gate 2000 and to recognize that the user is not authorized to pass if the user is not a registered user.

  The authentication device 1000 may be configured to be provided for all the proximity detection sensors 200 and the imaging devices 300. However, depending on the number of users, the authentication devices 1000 may include the proximity detection sensors 200 and the imaging devices 300 in order to distribute the processing load. One may be provided for one set. When one authentication device 1000 is provided for one set of the proximity detection sensor 200 and the imaging device 300, the authentication system 1 is provided in each gate 2000. Since the authentication system 1 is provided for each gate 2000, a plurality of users can be simultaneously authenticated.

In this embodiment, an image of the skin from the neck to the chin of the passing user is used as the physical feature. As shown in FIG. 6A, the proximity detection sensor 200 and the imaging device 300 are disposed above the gate 2000. The imaging apparatus 300 is provided with an imaging range 2006 so as to look up at the user from the side, and images the periphery of the jaw from the user's neck.
Note that it is also possible to change the arrangement of the imaging device 300 and use it as a physical feature by imaging a hand rather than from the user's neck to chin. However, the hand may not be photographed at a predetermined position so that the user can freely move the hand. Moreover, it is not certain which side of the palm or the back of the hand can be imaged by the imaging apparatus 300. Therefore, in order to use it as a physical feature used for authentication, a certain amount of user cooperation is required, such as having the user hold his hand.

  The external output unit 1600 of the authentication apparatus 1000 outputs the determination result by the collation unit 1500. If the user is a registered user, the external output unit 1600 prevents the user from passing through. When the user is not a registered user, the external output unit 1600 prevents the user from passing. For example, the gate 2000 includes a movable arm. When the user is not a registered user, the external output unit 1600 drives the movable arm driving device to close the arm. Alternatively, the external output unit 1600 may notify the surroundings that a user who is not registered by causing the speaker to make a sound is about to pass through the gate 2000. The external output unit 1600 may simultaneously perform arm driving and speaker notification.

  In addition to the effects of the authentication system 1 of the first embodiment, the authentication system 1 of the third embodiment as described above can perform authentication processing without being noticed by the user even if there are many people.

  The present invention supplies a program that realizes one or more functions of the above-described embodiments to a system or apparatus via a network or a storage medium, and one or more processors in a computer of the system or apparatus read and execute the program This process can be realized. It can also be realized by a circuit (for example, ASIC) that realizes one or more functions.

Claims (15)

Image acquisition means for acquiring a captured image of a person;
Recording means for recording a registered image that is an image of a person to be registered;
Collation area determination means for determining a collation area for collating a part of the captured image with a part of the registered image;
Collating means for collating the image of the collation area of the captured image with the image of the collation area of the registered image to determine whether or not the person of the captured image is a registered person. And
Authentication device. The collation area determining means determines the same position where the skin is exposed in the captured image and the registered image, respectively, as the collation area.
The authentication device according to claim 1. The recording means records a registered image of each of a plurality of persons,
The collation area determination means determines a different position in the collation area according to a registered person,
The authentication apparatus according to claim 1 or 2. The collation area determination means determines a different position in the collation area for each collation in the registered image of the same person,
The authentication device according to claim 1. The collation area determination unit extracts a plurality of candidate areas that are candidates for the collation area from each of the captured image and the registered image, and a plurality of candidate areas of the captured image and a plurality of candidate areas of the registered image And determining the matching region according to a candidate region whose position overlaps,
The authentication apparatus of any one of Claims 1-4. The collation area determination means determines the collation area by finely adjusting the position of a candidate area where the positions overlap based on skin characteristics.
The authentication device according to claim 5. The recording means records the registered image and positional information of the plurality of candidate areas in the registered image in association with each other.
The authentication device according to claim 5 or 6. The recording means records the registered image and the positional information of the plurality of candidate areas in a plurality of physically different storage devices and records them,
The authentication device according to claim 7. An authentication system comprising: an imaging device that images a person; and an authentication device that authenticates the person,
The authentication device
Image acquisition means for acquiring a captured image obtained by imaging the person from the imaging device;
Recording means for recording a registered image that is an image of a person to be registered;
Collation area determination means for determining a collation area for collating a part of the captured image with a part of the registered image;
Collating means for collating the image of the collation area of the captured image with the image of the collation area of the registered image to determine whether or not the person of the captured image is a registered person. And
Authentication system. A sensor for detecting the person;
The authentication apparatus further includes a detection unit that acquires a detection result of the sensor,
The image acquisition unit acquires the captured image from the imaging device when the detection unit acquires a detection result indicating that the person is detected from the sensor.
The authentication system according to claim 9. Further comprising a projector for irradiating the person with light,
The image pickup device picks up an image of a range irradiated with light from the projector.
The authentication system according to claim 9 or 10. The imaging device performs imaging in one of visible light, near infrared, and a wavelength band between red and near infrared,
The authentication system of any one of Claims 9-11. A method executed by an authentication device that acquires a captured image obtained by imaging the person from an imaging device that images the person and authenticates the person,
The authentication device is
Record a registered image, which is an image of a person to be registered, in a predetermined recording means in advance,
Determined as a collation area for collating part of the captured image and part of the registered image,
It is determined whether or not the person of the captured image is a registered person by comparing the image of the verification area of the captured image with the image of the verification area of the registered image,
Authentication method. A computer connected to an imaging device for imaging a person,
Image acquisition means for acquiring a captured image obtained by imaging the person from the imaging device;
Recording means for recording a registered image that is an image of a person to be registered;
Collation area determination means for determining a collation area for collating a part of the captured image with a part of the registered image;
Collation means for collating the image of the collation area of the captured image with the image of the collation area of the registered image to determine whether the person of the captured image is a registered person;
Computer program to function as.   A computer-readable storage medium storing the computer program according to claim 14.

Images