JP2008071172A - Face authentication system, face authentication method, and access control device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a face authentication system, a face authentication method, and an access control device which allows a person to be authenticated to check whether or not he/she is likely to be out of the scope and improve face authentication performance by preventing the person from being out of the scope. <P>SOLUTION: When the face authentication system authenticates a pedestrian, it is determined whether or not the pedestrian is a preregistered person by comparing a face image acquired from the pedestrian with a preregistered dictionary information. When the face of the pedestrian is likely to be out of the scope of a camera, an end of a zoomed display image is highlighted regarding a side where the face is likely to be out of the scope. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention, for example, uses a pedestrian as an authentication target, and compares the face image acquired from the pedestrian with pre-registered dictionary information to determine whether the pedestrian is a pre-registered person. The present invention relates to an apparatus and a face authentication method.

  The present invention also relates to an entrance / exit management apparatus that manages entrance / exit for an entrance / exit area such as a room or facility that requires security using the face authentication apparatus.

  In general, an entrance / exit management device using a face authentication device has a camera such as a video camera, and the person to be authenticated stops in front of the camera and faces the lens of the camera. An image is captured and input by a camera, and the person is a pre-registered person by collating face characteristic information unique to the person obtained from the input face image with dictionary information registered in advance. If the person is registered in advance, the door of the entrance / exit area (room, facility, etc.) is opened (for example, see Patent Document 1).

  In this type of entrance / exit management device (face authentication device), as described above, since the person to be authenticated stops in front of the camera, the person's face is imaged. Is a pedestrian (moving person), and when it is used for the purpose of performing face authentication before the pedestrian approaches the door (specific position), imaging the face of such a walking person is There is a problem that you can not.

In view of this, it is considered that a camera and a display unit are arranged in the vicinity of the entrance of the entrance / exit area to perform face authentication of a pedestrian moving toward the entrance door. This is because a face image of a pedestrian moving toward the entrance door is captured and input by a camera, and the facial feature information unique to the person obtained from the input face image is registered in advance as dictionary information. By checking, it is determined whether the pedestrian is a pre-registered person. If the pedestrian is a pre-registered person, the entrance door is opened to allow access, otherwise the entrance door is opened. It is closed to prevent traffic.
In this way, by completing the face authentication process before the pedestrian reaches the entrance door, the pedestrian registered in advance can pass through the entrance without stopping.

  In the face authentication process for determining whether or not a person is registered in advance based on a face image, it is important in terms of improving the accuracy of face authentication to stably capture the front face. Therefore, a display unit is arranged in the vicinity of the camera, the pedestrian's face captured by the camera is displayed on the display unit with zoom, and the pedestrian is guided to look at the display unit, so that the front face is displayed. Techniques for shooting are also being considered.

  As a technique for acquiring the front face of a pedestrian, a method for determining the face direction from a face image obtained by imaging with a camera is known (for example, see Patent Document 2). This calculates the left and right contours of the left eye, right eye, and face from the input image, and then calculates the ratio between the distance between the left eye and the left contour of the face and the distance between the right eye and the right contour section of the face. Determine the face orientation. If the two distances are substantially equal, it is determined that the face is facing the front, and if one of the two distances is short, it is determined that the face is facing in that direction.

As a technique for detecting a face area from an input image, a technique for detecting a human face from an input image series from a camera is known (see, for example, Patent Document 3). In this method, face pattern detection is performed every few frames for the entire input image, and face pattern detection is performed only in the vicinity of the area in which the face is detected in subsequent frames. This is because the face pattern detection is not performed on the entire input image every frame, but when a face is detected once, the processing speed is improved by limiting the target area of the face pattern detection in subsequent frames. Based on this idea.
JP 2001-266152 A JP 2005-196567 A JP 2004-171490 A

  When a person is detected and the face of the person is zoomed on the display, it is not known where the person is in the camera's field of view. It is to come out. This causes a person to go out of the field of view of the camera before the number of face images necessary for face authentication is detected, leading to a decrease in authentication performance.

  Further, the face orientation determination in Patent Document 2 is used for the purpose of prompting a person to turn his face to the camera when the face of the pedestrian is not in front. This technology is based on the premise that face parts, specifically left and right eyes and left and right outlines of faces, are extracted first for face orientation determination, and when face illumination is dark and detection of the face parts fails. There is a problem that the face orientation cannot be determined. As a result, the possibility that a pedestrian's front face can be obtained is reduced, leading to a reduction in face authentication performance.

  Further, in the face area detection of Patent Document 3, since the face detection of the entire image is performed every several frames, detection of the face of a person appearing between the frames is delayed. As a result, when the technique of Patent Document 3 is used for the purpose of detecting the face of a pedestrian in the present invention, the number of faces extracted from a person is reduced, and the number of face images necessary for face authentication cannot be extracted. Can occur. Accordingly, there is a problem that the face authentication performance is deteriorated.

  Therefore, the present invention makes it possible to confirm whether or not the person to be authenticated is out of the imaging field of view, and to prevent the person from being out of the imaging field of view, thereby improving the face authentication performance. An object is to provide an authentication device, a face authentication method, and an entrance / exit management device.

  It is another object of the present invention to provide a face authentication device, a face authentication method, and an entrance / exit management device that make it easier to obtain a person's front face and consequently improve face authentication performance.

  Furthermore, the present invention provides a face authentication device, a face authentication method, and an entrance / exit management device in which face detection delay is eliminated, and as a result, a face image necessary for authentication can be easily obtained and face authentication performance is improved. With the goal.

  The face authentication device of the present invention is a face authentication device that performs face authentication of a person moving toward a specific position, and is provided between the moving person and the specific position, and at least of the moving person Image capturing means for capturing an image including a face, face detection means for detecting the face area of the person from the image captured by the image capturing means, and an image of the face area detected by the face detection means Authentication means for determining whether or not the person is a registered person by collating with dictionary information, and provided in the vicinity of the specific position and detected by the face detection means for the moving person Display means for enlarging and displaying an image of the face area, and detecting that the image of the face area displayed by the display means is likely to deviate from the imaging field of view of the imaging means; And comprising a display control means for controlling said display means so as to highlight the end of the enlarged display image on the side was almost off the.

  The face authentication device of the present invention is a face authentication device that performs face authentication of a person moving toward a specific position, and is provided between the moving person and a specific position, and the moving person Imaging means for capturing an image including at least a face, face detection means for detecting the face area of the person from the image captured by the imaging means, and an image of the face area detected by the face detection means registered in advance Authentication means for determining whether the person is a person registered in advance by collating with the dictionary information, and the face detection means includes a plurality of images captured by the imaging means. The difference between the images is calculated to detect the change area, the optical flow extraction process is performed on the obtained change area, and the obtained optical flow is close in position and the flow is detected. The grouping process is performed so that the optical flows with the same direction and size are in the same group, and the peripheral features are extracted for the group in which the number of optical flows constituting the obtained group is larger than the threshold, and the obtained peripheral features A range in which the value of the first derivative sandwiched between positions where the first derivative of the quantity has a positive value and a negative value is close to “0” is extracted as a head candidate for the person, and face pattern detection is performed on the obtained head candidate. This is characterized in that the face area is detected.

  The face authentication device of the present invention is a face authentication device that performs face authentication of a person moving toward a specific position, and is provided between the moving person and a specific position, and the moving person Imaging means that captures an image including at least a face, a face detection means that detects the face area of the person from the image captured by the imaging means, and an image of the face area detected by the face detection means The person is registered in advance by collating the face direction determining means for determining whether or not the face is determined and the image of the face area determined to be front facing by the face direction determining means with dictionary information registered in advance. Authentication means for determining whether the person is a person, and the face orientation determination means has a height shorter than the height and width of the circumscribed rectangle of the face area detected by the face detection means. The long When the value divided by the length of the face is greater than or equal to the threshold value, the face is determined not to face the front, and when the value obtained by dividing the shorter length by the longer length is less than or equal to the threshold value, the face pattern Detection is performed, the similarity with a face detection dictionary pattern created in advance from a face facing forward is calculated, and when the obtained similarity is equal to or less than a threshold, it is determined that the face is not facing forward, and the obtained When the similarity is equal to or greater than a threshold, it is determined that the face is facing the front.

  Further, the face authentication method of the present invention is a face authentication method for performing face authentication of a person moving toward a specific position, by an imaging means provided between the moving person and the specific position, An imaging step of capturing an image including at least the face of the moving person, a face detection step of detecting the face region of the person from the image captured by the imaging step, and a detection of the face region detected by the face detection step The person who moves by the authentication step of determining whether the person is a person registered in advance by comparing the image with the dictionary information registered in advance and the display means provided in the vicinity of the specific position A display step for enlarging and displaying an image of the face area detected by the face detection step, and an image of the face area displayed by the display step is out of the imaging field of view of the imaging means. Detects that became so, and a display control step for controlling the display means so as to highlight the end of the enlarged display image on the side was almost out from the imaging field.

  Further, the face authentication method of the present invention is a face authentication method for performing face authentication of a person moving toward a specific position, by an imaging means provided between the moving person and the specific position, An imaging step of capturing an image including at least the face of the moving person, a face detection step of detecting the face region of the person from the image captured by the imaging step, and a detection of the face region detected by the face detection step An authentication step of determining whether the person is a pre-registered person by comparing an image with pre-registered dictionary information, and the face detection step is imaged by the imaging step. The difference between images is calculated for a plurality of images to detect a change area, optical flow extraction processing is performed on the obtained change area, and the position of the obtained optical flow is determined. In addition, grouping processing is performed so that optical flows whose flow direction and size are close to each other are in the same group, and peripheral features are extracted for a group in which the number of optical flows constituting the obtained group is larger than a threshold value. A range in which the value of the first derivative sandwiched between positions where the first derivative of the obtained peripheral feature value has positive and negative values is close to “0” is extracted as a candidate for the person's head, and the obtained head candidate A face area is detected by performing face pattern detection for.

  Further, the face authentication method of the present invention is a face authentication method for performing face authentication of a person moving toward a specific position, by an imaging means provided between the moving person and the specific position, An imaging step of capturing an image including at least the face of the moving person, a face detection step of detecting the face region of the person from the image captured by the imaging step, and a detection of the face region detected by the face detection step A face orientation determination step for determining whether or not the image is front-facing, and an image of the face area determined to be front-facing by the face orientation determination step and the dictionary information registered in advance are compared. An authentication process for determining whether the person is a registered person, and the face orientation determination process is performed by a height and a width of a circumscribed rectangle of the face area detected by the face detection process. When the value obtained by dividing the length of one side by the length of the longer side is equal to or greater than the threshold value, the face is determined not to face front, and when the value obtained by dividing the length of the shorter side by the length of the longer side is equal to or less than the threshold value Face pattern detection is performed on the face area, the degree of similarity with a face detection dictionary pattern created in advance from a face facing forward is calculated, and when the obtained degree of similarity is equal to or less than a threshold value, the face is not facing forward And determining that the face is facing forward when the obtained similarity is equal to or greater than a threshold value.

  The entrance / exit management device of the present invention performs face authentication of a person moving toward an entrance / exit gate provided in an entrance / exit target area, and controls opening / closing of the entrance / exit gate based on the authentication result. An exit management apparatus, which is provided between the moving person and the entrance / exit gate, picks up an image including at least a face of the moving person, and an image picked up by the image pickup means Whether the person is a pre-registered person by collating the face detection means for detecting the face area of the person and the image of the face area detected by the face detection means with previously registered dictionary information. Authentication means for determining, gate control means for controlling opening / closing of the entrance / exit gate based on a determination result of the authentication means, provided near the entrance / exit gate, and moving Display means for enlarging and displaying the image of the face area detected by the face detection means for a person, and the image of the face area displayed by the display means is likely to be out of the imaging field of view of the imaging means. And a display control means for controlling the display means to highlight the end of the enlarged display image on the side that is likely to deviate from the imaging field of view.

  The entrance / exit management device of the present invention performs face authentication of a person moving toward an entrance / exit gate provided in an entrance / exit target area, and controls opening / closing of the entrance / exit gate based on the authentication result. An exit management apparatus, which is provided between the moving person and the entrance / exit gate, picks up an image including at least a face of the moving person, and an image picked up by the image pickup means Whether the person is a pre-registered person by collating the face detection means for detecting the face area of the person and the image of the face area detected by the face detection means with previously registered dictionary information. Authentication means for determining, and gate control means for controlling opening / closing of the entrance / exit gate based on the determination result of the authentication means, wherein the face detecting means is the imaging means The change region is detected by calculating the difference between the images for a plurality of captured images, and the optical flow extraction process is performed on the obtained change region. The grouping process is performed so that the optical flows whose directions and sizes are close to each other are in the same group, and the peripheral features are extracted from the group in which the number of optical flows that make up the obtained group is larger than the threshold, and the obtained peripheral feature amount A range in which the value of the first derivative sandwiched between positions where the first derivative has positive and negative values is close to “0” is extracted as a head candidate for the person, and face pattern detection is performed on the obtained head candidate. Thus, the face area is detected.

  Furthermore, the entrance / exit management device of the present invention performs face authentication of a person moving toward an entrance / exit gate provided in an entrance / exit target area, and controls the entrance / exit gate based on the authentication result. An exit management apparatus, which is provided between the moving person and the entrance / exit gate, picks up an image including at least a face of the moving person, and an image picked up by the image pickup means Face detection means for detecting the face area of the person, face orientation determination means for determining whether the image of the face area detected by the face detection means is front-facing, and front-facing by the face orientation determination means Authentication means for determining whether or not the person is a pre-registered person by comparing the image of the face area determined to be a pre-registered dictionary information, and the authentication means Gate control means for controlling opening / closing of the entrance / exit gate based on the determination result, and the face orientation determination means has a height and width of a circumscribed rectangle of the face area detected by the face detection means, When the value obtained by dividing the shorter length by the longer length is equal to or greater than the threshold value, it is determined that the face is not facing the front, and the value obtained by dividing the shorter length by the longer length is equal to or less than the threshold value. When a face pattern is detected for the face area, a similarity with a face detection dictionary pattern created in advance from a face facing forward is calculated. When the obtained similarity is equal to or less than a threshold, the face is not facing front And when the obtained similarity is equal to or greater than a threshold value, it is determined that the face is facing forward.

  According to the present invention, it becomes possible for a person to be authenticated to check whether or not he / she is outside the imaging field of view, and a face whose face authentication performance is improved by taking care not to deviate from the imaging field of view. An authentication device, a face authentication method, and an entrance / exit management device can be provided.

  Furthermore, according to the present invention, it is possible to provide a face authentication device, a face authentication method, and an entrance / exit management device that make it easier to obtain a person's front face and consequently improve face authentication performance.

  Furthermore, according to the present invention, there is provided a face authentication device, a face authentication method, and an entrance / exit management device in which face detection delay is eliminated, and as a result, a face image necessary for authentication can be easily obtained and face authentication performance is improved. it can.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
First, the outline of the present invention will be briefly described. In the present invention, for example, as shown in FIG. 2, a door (entrance / exit gate) provided at the entrance 3 of an entrance / exit area (room, facility, etc.) 2 by moving the traffic band 1 in the direction of the arrow e shown in the figure. The face of a pedestrian (hereinafter also referred to as a person) M facing 4 is imaged with a camera. Specifically, while the pedestrian M is present between the point d and the point b in the traffic zone 1, an image including at least a heel of the pedestrian M is captured with a camera, and the point a near the entrance 3 from the point b. While the pedestrian M arrives, it is determined from the acquired image whether the pedestrian M is a person registered in advance. When it is determined that the person is registered in advance, the door 4 is unlocked.

  FIG. 1 schematically shows a configuration of an entrance / exit management apparatus to which the face authentication apparatus according to the present embodiment is applied. The entrance / exit management device includes a video camera (hereinafter simply referred to as a camera) 11 that captures an image including at least a face of the pedestrian M, and a face area of the pedestrian M from an image captured by the camera 11. Detection and tracking unit 12 as a face detection means for detecting and tracking a face detection dictionary pattern storage unit (memory) 13 for storing a face detection dictionary pattern used for detection of a face region, and for a pedestrian M who is tracking Face orientation determination unit 14 as a face orientation determination unit that determines whether or not the face direction is facing the front of the camera 11. Face authentication as a dictionary storage unit in which a plurality of dictionary information is registered (stored) in advance. By comparing the image (face image) of the face area obtained by the dictionary (memory) 15 and the face detection / tracking unit 12 with dictionary information registered in the face authentication dictionary 15 in advance, the pedestrian M A face authentication unit 16 as an authentication unit for determining whether or not a person is registered in advance, a gate control unit 17 as a gate control unit for controlling the door 4 based on a determination result of the face authentication unit 16, and a pedestrian M On the other hand, the display unit 18 as a display means for displaying the current face authentication status, the speaker 19 as a guidance means for guiding the pedestrian M by voice, and the display authentication as a control means for overall control. A control unit 20 is included.

Hereinafter, each component will be described in detail.
For example, as shown in FIG. 2, the display unit 18 is installed in the vicinity of the entrance 3 on the side of the traffic zone 1. The installation height of the display unit 18 is, for example, about the average value of the height of the pedestrian M.
The display unit 18 displays the current face authentication status and the like for the pedestrian M. Specifically, when the pedestrian M cannot detect the face far from the door 4 or when the pedestrian M starts detecting, the entire image 31 obtained from the camera 11 as shown in FIG. Is displayed. A frame 32 in FIG. 3A indicates a detected face area.

  When a face is detected and the pedestrian M approaches a certain distance from the door 4, more specifically, when approaching a point c in FIG. 2, as shown in FIG. 3B, the detected face Enlarge and display the image.

  Further, when the pedestrian M approaches the door 4, specifically, when approaching the point b in FIG. 2, it is determined whether or not the person is a registered person from the image obtained from the camera 11. Judgment (face authentication) is started, and face authentication is completed before reaching point a. At this time, as shown in FIG. 3C, a guidance sentence “authenticating” is displayed.

  When authentication is completed and it is determined that the person is registered in advance, a guidance message indicating that entry is permitted is displayed to the pedestrian M as shown in FIG. To do. Otherwise, as shown in FIG. 3 (e), a guidance sentence indicating that entry has been refused is displayed to the pedestrian M.

The camera 11 captures a face image of the pedestrian M moving toward the door 4 and, for example, a moving image sequence of a person approaching the camera 11 as shown in FIGS. For example, as shown in FIG. 2, it is installed in the vicinity of the display unit 18. Reference numeral 11 a denotes an imaging field of view of the camera 11.
The speaker 19 provides various guidance by voice to the pedestrian M, and is installed in the vicinity of the camera 11, for example, as shown in FIG.

  The face detection / tracking unit 12 detects an area where a face exists as a face candidate area from a moving image sequence of a person approaching the camera 11 as shown in FIG. The detected face candidate areas are temporally correlated to obtain the face candidate areas as a time series. Here, the latest input image from the camera 11 is called the current frame image, and the image obtained by the previous image input is called the previous frame image. In addition, the time when the image of the current frame is obtained is represented as t, and the time when the image of the previous frame is obtained is represented as t-1.

  FIG. 5 shows a processing flow of the face detection tracking unit 12. First, in the image input in step S1, an image is input from the camera 11 at regular intervals. Next, in the face initial detection in step S2, a region having a change is extracted from the input image series, and a region where a face exists is detected as a face candidate region from the extracted change region.

  Next, in the face tracking process in step S3, the face candidate area obtained by the initial face detection is tracked. Specifically, the position where the face candidate region in the previous frame is present in the image of the current frame is determined. After the face tracking process, the face candidate area in the current frame is sent to the display authentication control unit 20 (step S4). The display authentication control unit 20 switches the display according to this state.

  If a certain number of face images (M_IMG) or more are collected for the person whose face is being tracked (step S5), the obtained face images are sent to the display authentication control unit 20 (step S6). The display authentication control unit 20 performs face authentication processing using the obtained face image.

  Here, the face initial detection process in step S2 will be described in detail. In the face initial detection process, an area having a change is extracted from the input image series, and a face pattern is detected from the extracted change area area. FIG. 6 shows the flow of the face initial detection process.

The input of the face initial detection process is an image sequence as shown in FIG. First, the inter-frame difference in step S11 is a process of comparing two consecutive frame images and extracting a pixel having a change. When the input image in the previous frame is It-1 (x, y) and the input image in the current frame is It (x, y), the inter-frame difference image DIFFt (x, y) is calculated as Is done.

  Here, assuming that the area where DIFFt (x, y) = 1 is a change area, for example, the outline of a person is extracted as the change area as indicated by reference numeral A in FIG. At the same time, background noise and the like are also extracted as indicated by reference numeral B in FIG. In FIG. 7, pixels with DIFFt (x, y) = 1 are represented in black, and pixels with DIFFt (x, y) = 0 are represented in white.

  As an inter-frame difference calculation method, the following incremental code inter-frame difference may be used. Specifically, horizontal and vertical incremental code images DIFF_S_Xt-1 (x, y) and DIFF_S_Yt-1 (x, y) for time t-1 and horizontal and vertical increments for time t. Coded images DIFF_S_Xt (x, y) and DIFF_S_Yt (x, y) are calculated.

DIFF_S_Xt (x, y) and DIFF_S_Yt (x, y) are the following images.

Further, the incremental code inter-frame difference image DIF_St (x, y) is as follows.

  As the inter-frame difference, either the luminance inter-frame difference DIFFt (x, y) or the incremental code inter-frame difference DIFF_St (x, y) may be used. The luminance frame difference DIFFt (x, y) has a small amount of calculation, but has a property that it is weak against luminance fluctuation because the luminance difference threshold T_DIFF is used. Incremental code inter-frame difference DIFF_St (x, y) is more computationally intensive than luminance inter-frame difference DIFFt (x, y), but is less susceptible to luminance fluctuations because it does not use a luminance difference threshold. There is a nature.

  Next, in step S12, the optical flow extraction divides the input image of the previous frame into small blocks having a fixed size of 2 × w on a side. Here, w is a constant. Among the divided small blocks, an optical flow from the previous frame to the current frame is calculated for a block including a pixel that is determined to have a change area due to an inter-frame difference. For example, when the change region is as shown in FIG. 7, the optical flow is calculated for the thick block BL shown in FIG.

When the small block of the previous frame is, for example, as shown in FIG. 9, the optical flow is calculated using the rectangular area of the current frame whose side is 2 × SW in FIG. 9 as a search range. Specifically, the pixel at the center of the i-th block at time t−1 is

As the value of SAD

The value of (dx, dy) that minimizes is obtained. Pixel at time t-1

Pixel at time t

Is determined to correspond.

  An example in which the optical flow is calculated for each block is shown in FIG. In FIG. 10, the optical flow is shown as a line connecting corresponding pixels in the previous frame and the current frame on the image of the previous frame.

  Next, the head candidate region extraction in step S13 divides the obtained optical flows into groups so that those close to each other and similar in movement amount belong to the same group. FIG. 11 shows an example in which the optical flow is divided into groups. If the number of optical flows constituting the obtained group is equal to or less than the threshold value, it is deleted. For example, when the threshold value is “2”, background noises C, D, E, F, and G in FIG. 11 are deleted.

  Next, the peripheral feature PE (x) is calculated for the remaining group. Specifically, the input image is scanned in the y-axis direction from the point (x, 0). Let PE (x) be the minimum distance to the optical flow on the scanned line. If there is no optical flow present on the scanned line, PE (x) is the y size (height) of the input image. FIG. 12 shows a specific example of the peripheral feature.

Next, as shown in FIG. 13, the first derivative of the peripheral feature PE (x) is defined as PE ′ (x). In the first derivative PE (x), the left side is a negative edge and the right side is a positive edge, and their absolute values are larger than the threshold value and are sandwiched between them.

The range which becomes is the vertex candidate H. For example, in the case of FIG. 13, the I portion is the leftmost negative edge, the J portion is the right positive edge, and the K portion is sandwiched between them.

Range.

  Next, a head candidate region is extracted from the obtained top candidate. The head candidate area is a rectangle whose height is a constant multiple (A_TOP) of the width (W_TOP) of the top candidate, with the top candidate area as the upper side. FIG. 14 shows a specific example of the head candidate area.

  Next, in the face pattern detection in step S14, a face pattern is detected for the obtained head candidate region. However, those that have already been extracted in the face initial detection process and are tracked in the face tracking process are excluded.

  Examples of face pattern detection methods include literature (Mita, Kaneko, Hori, “Proposal of Spatial Difference Probability Template Suitable for Authentication of Images Containing Minute Differences”, Proceedings of the 9th Meeting of Image Sensing Symposium, SSII03, The technique described in 2003) can be used. This is a technique in which a face detection dictionary pattern is created in advance from a face learning pattern, and an area having a high similarity with the face detection dictionary pattern is searched from an input image.

  For example, the face detection dictionary pattern is created from a large number of face images facing the front. The face detection dictionary pattern created in this way is stored in the face detection dictionary pattern storage unit 13 and is used when detecting the face pattern, so that a face candidate region facing almost front is detected. The detected face candidate area is represented as a square face candidate rectangle as shown in FIG. In FIG. 15, a symbol L indicates a head candidate region, and a symbol N indicates a face candidate region.

  Next, the face tracking process in step S3 will be described in detail. In the face tracking process, the face candidate area in the current frame is obtained from the face candidate area in the previous frame. FIG. 16 shows the flow of the face tracking process.

  First, an optical flow is extracted for the face candidate area of the previous frame (step S21). Specifically, the face candidate area of the previous frame is divided into small blocks, and the position of each block in the current frame is calculated using template matching or the like. FIG. 17 shows an example of the optical flow of the face candidate area.

  The purpose of the optical flow extraction in the face initial detection process (step S2) described above is to roughly narrow the change region. On the other hand, in the optical flow extraction in the face tracking process (step S3), since the purpose is to calculate the movement of the face, the accuracy is required more than the optical flow extraction in the face initial detection process.

  Therefore, in the extraction of the optical flow in the face tracking process, a small block division method different from the optical flow extraction process in the face initial detection process may be used. For example, in the calculation of the optical flow in the face tracking process, in order to extract more optical flows, the small blocks may be arranged so that the adjacent small blocks overlap each other. This improves the accuracy of the face tracking process.

Here, the center point of the small block in the face candidate area of the previous frame is

And the center point

The point of the current frame corresponding to

And

Next, face deformation parameter calculation is performed (step S22). this is,

The conversion formula to be transferred to is calculated. For example, as a conversion formula:

Is used. This is a conversion formula for scaling and translation of the x-axis and y-axis, respectively.

Next, Ax, Bx, Ay, By is predicted using regression analysis, and the obtained values are set as A′x, B′x, A′y, B′y.

If both values of Expression 15 are smaller than a predetermined threshold, calculation of face deformation parameters is obtained (step S23), and the circumscribed rectangle R ^t-1 of the face candidate area of the previous frame is used to calculate the current frame. calculate the circumscribed rectangle R ^t of the face candidate region, and the position of the face candidate region of the current frame (step S24), and ends the process.

R ^t-1 of the upper left and the coordinates of the points of the lower right, ^{respectively, (LEFT t-1, TOP} t-1), (RIGHT t-1, and BOTTOM ^t-1, of the upper left point and the lower right of the ^{R t} If the coordinates are (LEFT ^t , TOP ^t ) and (RIGHT ^t , BOTTOM ^t ), respectively,

Is converted as follows.

  If the calculation of the deformation parameter has failed (step S23), that is, if both of the values of Formula 15 are equal to or greater than a predetermined threshold, face pattern detection is performed for an area obtained by extending the face candidate area of the previous frame by a certain size. (Step S25). If a face candidate area is detected from the face pattern (step S26), it is set as a face candidate area in the current frame (step S27), and the process is terminated. If the face candidate area is not detected by the face pattern (step S26), the face candidate area of the previous frame is set as the face candidate area of the current frame (step S28).

  Next, it is checked whether or not the number of frames in which the face candidate area of the previous frame is continuously the face candidate area of the current frame is greater than or equal to a predetermined threshold T_NOT_DETECTED (step S29), and if not greater than the threshold T_NOT_DETECTED, the process ends. If it is equal to or greater than the threshold value T_NOT_DETECTED, it is assumed that the person is no longer in the field of view of the camera, face tracking is stopped (step S30), and the process is terminated.

  The face orientation determination unit 14 is a process of determining whether or not the face orientation being tracked is facing the front with respect to the camera 11. FIG. 18 shows an example of a circumscribed rectangle of the face direction and the face candidate area. The face area detected in the initial face detection is a face facing front as shown in FIG. This is because a face-oriented dictionary is used as the face pattern dictionary. A square rectangle as shown in FIG. 18E is given to the detected front-facing face, and tracking is performed as a face candidate area. If the face is front-facing as shown in FIG. 18E, the face remains square, and the width and height of the circumscribed rectangle are approximately equal. However, as shown in FIGS. 18D and 18F, the height becomes larger than the width of the circumscribed rectangle. Similarly, as shown in FIGS. 19B and 19H, the width becomes longer with respect to the height of the circumscribed rectangle when facing upward or downward.

Accordingly, the face orientation determination unit 14 performs processing according to the flowchart shown in FIG. First, the orientation of the face is determined based on the ratio between the length and width of the circumscribed rectangle of the face candidate area. Specifically, among the width and height of the circumscribed rectangle of the face candidate region, the longer length is len_long and the shorter length is len_short (step S41).

(Step S42), it is determined that the face is not facing the front (Step S43), and the process is terminated. Here, T_len is a threshold value larger than “1”.

18 is not satisfied, that is, when the aspect ratio of the circumscribed rectangle of the face candidate region is close to “1”, not only when the face is facing the front as shown in FIG. In some cases, it may face diagonally upward or diagonally downward as in (a), (c), (g), and (i).

Therefore,

If this is not true (step S42), face pattern detection is applied to the face candidate area, and the similarity to the face detection dictionary pattern is calculated (step S44). When the obtained similarity is smaller than the predetermined threshold T_sim (step S45), it is determined that the face direction is not the front (step S43), and the process is terminated. When the obtained similarity is larger than the predetermined threshold T_sim (step S45), it is determined that the face is facing the front (step S46), and the process ends.

  The face authentication unit 16 determines whether the face image obtained by the face detection tracking unit 12 is a person registered in the face authentication dictionary 15 in advance. Specifically, for example, as described in Japanese Patent Application Laid-Open No. 2001-266152, a registrant (pedestrian) face image is prepared in advance, and unique feature information extracted from the face image is used as dictionary information. It is stored (registered) in the face recognition dictionary 15. Then, the face area image (feature information) obtained by the face detection / tracking unit 12 and the dictionary information registered in advance in the face recognition dictionary 15 are collated to obtain a similarity between them, and the obtained similarity is set in advance. If it is equal to or greater than the threshold value, it is determined that the pedestrian M is a pre-registered person. If the obtained similarity is not equal to or greater than the threshold value, the pedestrian M is determined not to be a pre-registered person.

  When the object to be controlled is the door 4 as shown in FIG. 2, the gate control unit 17 unlocks when the passable signal from the display authentication control unit 20 is on, and locks when the passable signal is off. To. In addition, when the object to be controlled is a gate device, a warning sound is emitted when it passes through the gate device even when the pass signal is off, or the flapper provided in the gate device is closed and Or stop. Specifically, control is performed as shown in the flowchart of FIG.

  The display authentication control unit 20 controls the entire apparatus, and the process flow is shown in the flowchart of FIG. Hereinafter, the flow of processing will be described based on the flowchart of FIG.

  First, the processing result of the face detection / tracking unit 12 is acquired (step S51), and a face orientation guidance process (details will be described later) is performed (step S52). Next, it is checked whether or not the person to be authenticated (pedestrian M) is closer to the camera 11 than the point d in the traffic zone 1 (step S53), and the pedestrian M is closer to the camera 11 than the point d. If not, the image obtained from the camera 11 is displayed on the display unit 18 (step S54), the process returns to step S51, and the same operation as described above is repeated.

  In step S53, when the face area is detected by the face detection and tracking unit 12, and the pedestrian M from which the face area is detected is located closer to the camera 11 than the point d, the pedestrian M is set as a face authentication target. The display content of the display unit 18 is changed according to the location of the pedestrian M. The location of the pedestrian is predicted from the size of the detected face candidate area.

  That is, in step S53, when the pedestrian M from which the face area is detected is located closer to the camera 11 than the point d, whether the pedestrian M is between the points d and c in the traffic zone 1 or not. (Step S55), and when the pedestrian M is between the point d and the point c, the entire image 31 obtained from the camera 11 is displayed on the display unit 18 as shown in FIG. At the same time, the frame 32 indicating the face area detected at this time is displayed (step S56), and the process returns to step S51 to repeat the same operation as described above. Here, the input image is displayed with the left and right mirror images reversed.

  If the result of the check in step S55 is that the pedestrian M is not between the points d and c, it is checked whether or not the pedestrian M is between the points c and b in the traffic zone 1 (step S57). ) When the pedestrian M is between the points c and b, as shown in FIG. 3B, the detected face image is enlarged and displayed on the display unit 18 (step S58). Return and repeat the same operation as above. Here, the enlarged display (zoom display) of the face image is displayed with the left and right mirror images reversed.

  If the result of the check in step S55 is that the pedestrian M is not between the points c and b, the number of face-to-face images to be authenticated detected by the face detection and tracking unit 12 is the minimum number required for authentication. It is checked whether or not (N_IMG sheets) or more have been collected (step S59). As a result of this check, if a certain number or more of the images to be authenticated cannot be collected, a message “authentication NG” indicating that the authentication has failed is displayed on the display unit 18 as shown in FIG. S60), returning to step S51, the same operation as described above is repeated.

  As a result of the check in step S59, when a certain number or more of the authentication target saddle images can be collected, the face images other than those facing the front are removed from the face images obtained by the face detection tracking unit 12 (step S61). Only the face image facing the front is sent to the eyelid authentication unit 16, and the face authentication process is started (step S62). At this time, as shown in FIG. 3C, a message “authenticating” is displayed on the display unit 18 (step S63), and the process waits until the wrinkle authentication process is completed (step S64).

  顏 When the authentication process is completed, it is checked whether the authentication result is successful or unsuccessful (step S65). If the authentication result is successful, as shown in FIG. "Please" message is displayed (step S66), and a signal allowing passage to the gate control unit 17 is turned on for a predetermined time (step S67). Thereby, the pedestrian M can pass through the door 4.

  Next, it waits until a certain time elapses after turning on the passable signal, or until a pass completion signal of the pedestrian M is received from the gate control unit 17 (step S68). When a certain period of time elapses after the passable signal is turned on, or when a pass completion signal of the pedestrian M is received from the gate control unit 17, the process returns to step S51, and the same operation is repeated.

  As a result of the check in step S65, if the authentication result is unsuccessful, as shown in FIG. 3E, a message “authentication NG” is displayed on the display unit 18 for a certain period of time (step S69), and the process returns to step S51. The same operation as above is repeated.

  FIG. 22 shows the flow of the face direction guiding process in step S52, which will be described below. The face direction guidance process is performed according to the determination result of the face direction determination unit 14. That is, first, the determination result of the face direction determination unit 14 is acquired (step S71), and it is checked whether or not the pedestrian M has the front direction (step S72). As a result of this check, when it is determined that the face direction is the front, the process is terminated. When it is determined that the face direction is not the front, Please give a voice guidance such as “Please see” (step S73), and the process is terminated.

  FIG. 23 shows the flow of zoom display processing in step S58, which will be described below. First, an image (face image) of the detected face candidate area is cut out from the input image obtained from the camera 11 and displayed on the display unit 18 (step S81). Next, it is checked whether or not the head position of the displayed face image is close to the end of the input image (step S82). If the head position is not close to the end of the input image, the processing is performed. When the head position is close to the end of the input image, the end of the zoom display image is highlighted on the side close to the end of the input image (step S83), and the process ends.

As described above, in the zoom display process, when the face image of the pedestrian M is likely to protrude from the input image, that is, when the face image is likely to be out of the imaging field of view of the camera 11, it is detected, For example, as shown in FIGS. 24 and 25, the end portion is highlighted on the side that is likely to protrude.
In FIG. 24, since the face image is close to the left end in the input image, the end on the same side is highlighted in the zoom display (portion 41a). Since the zoom image is mirror-inverted, the right side is highlighted in FIG. In FIG. 25, since the face image is close to the upper end in the input image, the end (upper end) on the same side is highlighted in the zoom display image (the portion denoted by reference numeral 41b). 24 and 25, (a) shows an input image, and (b) shows a zoom display image corresponding to the input image.

  Here, as a method of highlighting the end of the zoom display image, for example, as shown in FIGS. 24 and 25, the highlight display portions 41a and 41b are formed in a strip shape and displayed in a color different from other portions. Or a method of blinking the highlighting portions 41a and 41b, or a method of displaying the highlighting portions 41a and 41b with pictorial symbols or the like. In short, it should be highlighted so that it can be distinguished from other parts.

  Thus, the outer frame (edge) is highlighted so that the position of the authentication target pedestrian M in the input image can be grasped by himself / herself. While the zoom image is displayed, only the face of the authentication target pedestrian M is shown, so that the pedestrian M can notice by the highlighted display that he / she is likely to be out of the field of view of the camera 11. Accordingly, the pedestrian M can move to a position where the face can be easily photographed by the camera 11 by slowing down the walking speed or approaching the camera 11 side.

  As described above, according to the above embodiment, when the pedestrian's face is likely to be out of the field of view of the camera, the end of the zoom display image is highlighted on the side that is likely to be out of the field of view. , Face recognition performance is improved. Specifically, the pedestrian to be authenticated can confirm whether or not he / she is going to go out of the field of view of the camera himself / herself. The number of face images required for face authentication can be obtained by changing the direction of walking or slowing down the walking speed so that it does not deviate from the camera's field of view. It becomes like this. Thereby, face authentication performance improves.

  In addition, the face detection dictionary pattern created from the face image facing in front is used for initial face detection, the face is tracked by optical flow, and the shorter length is set by the height and width of the circumscribed rectangle of the face candidate area. When the value divided by the longer length is greater than or equal to the threshold, or when the above value is less than or equal to the threshold and the similarity to the face detection dictionary pattern is less than or equal to the threshold, it is determined that the face is not facing front As a result, the accuracy of the face orientation determination is improved. Specifically, for example, even if the face direction cannot be determined by the method of determining the face direction after the illumination light is dark and the face part is detected, the present invention does not detect the face part. Face orientation can be determined. Thereby, face direction guidance function processing functions and it becomes easy to obtain a pedestrian's front face. As a result, face recognition performance is improved.

  Further, the inter-frame difference is calculated for the input image sequence to detect the change region, optical flow extraction is performed for the obtained change region, the obtained optical flow is close in position, and the flow direction and size are Grouping so that near optical flows are in the same group, and extracting the peripheral features for the group in which the number of optical flows constituting the obtained group is larger than the threshold, the first derivative of the obtained peripheral feature value is positive. And a negative first-order differential value sandwiched between positions having negative values is extracted as a candidate for the top of the head, so that there is no detection delay of the face area, and as a result, a face image necessary for face authentication The possibility that the number of images cannot be obtained is reduced and the face recognition performance is improved.

  In the present invention, since a pedestrian approaching the camera is a detection target, face pattern detection may be performed by narrowing down to a region where there is a change. Therefore, in the present invention, a change area is detected from the input image, a head candidate is further detected from the detected change area, and face pattern detection is performed for each frame only for the obtained head candidate area, thereby processing speed. The face pattern can be detected every frame without slowing down. This eliminates the face detection delay.

The block diagram which shows schematically the structure of the entrance / exit management apparatus with which the face authentication apparatus which concerns on embodiment of this invention is applied. The schematic diagram explaining the example of installation of an entrance / exit management apparatus. The schematic diagram which shows the example of a display screen in a display part. The schematic diagram which shows the example of the input image series input from a camera. The flowchart explaining the flow of a process of a face detection tracking part. The flowchart explaining the flow of a face initial detection process. The figure for demonstrating the difference between frames in a face initial detection process. The figure for demonstrating the calculation of the optical flow in a face initial detection process. The figure for demonstrating the calculation of the optical flow in a face initial detection process. The figure which shows the example of the optical flow in a face initial stage detection process. The figure for demonstrating grouping of the optical flow in a face initial detection process. The figure which shows the example of the peripheral characteristic in a face initial stage detection process. The figure for demonstrating the process which extracts a vertex candidate from a peripheral feature in a face initial detection process. The figure which shows the example of the head candidate in a face initial detection process. The figure which shows the example of the face candidate in a face initial detection process. The flowchart explaining the flow of a face tracking process. The figure for demonstrating the optical flow about a face candidate area | region. The flowchart explaining the flow of a process of a face direction determination part. The figure which shows the example of the circumscribed rectangle of face direction and a face candidate area | region. The flowchart explaining the flow of a process of a gate control part. The flowchart explaining the flow of a process of a display authentication control part. The flowchart explaining the flow of a face direction guidance process. The flowchart explaining the flow of a zoom display process. The schematic diagram which shows the 1st example of a zoom display image when a face is likely to become out of the visual field of a camera. The schematic diagram which shows the 2nd example of a zoom display image when a face is likely to become out of the visual field of a camera.

Explanation of symbols

  M ... Pedestrian, 1 ... Traffic band, 2 ... Entrance / exit area (room, facility, etc.), 3 ... Entrance, 4 ... Door (entrance / exit gate), 11 ... Video camera (imaging means), 12 ... Face detection Tracking unit (face detection unit), 13 ... Face detection dictionary pattern storage unit, 14 ... Face direction determination unit (face direction determination unit), 15 ... Face authentication dictionary (dictionary storage unit), 16 ... Face authentication unit (authentication unit) , 17... Gate control section (gate control means), 18... Display section (display means), 19. Speaker (guide means), 20... Display authentication control section (control means), 41 a and 41 b.

Claims (15)

A face authentication device for performing face authentication of a person moving toward a specific position,
An imaging unit that is provided between the moving person and a specific position and that captures an image including at least the face of the moving person;
Face detection means for detecting the face area of the person from the image captured by the imaging means;
Authentication means for determining whether the person is a pre-registered person by collating an image of the face area detected by the face detection means with pre-registered dictionary information;
Display means provided in the vicinity of the specific position and displaying an enlarged image of the face area detected by the face detection means for the moving person;
It is detected that the image of the face area displayed by the display means is likely to deviate from the imaging visual field of the imaging means, and the end of the enlarged display image is highlighted on the side that is likely to deviate from the imaging visual field. Display control means for controlling the display means,
A face authentication apparatus comprising:   The face detection means calculates inter-image differences for a plurality of images captured by the imaging means, detects a change area, performs optical flow extraction processing on the obtained change area, and obtains the obtained optical flow Grouping processing is performed so that optical flows that are close in position and close to the flow direction and size are in the same group. A range in which the first derivative value between the positions where the first derivative of the obtained peripheral feature value is between positive and negative values is close to “0” is extracted as a candidate for the top of the head. The face authentication apparatus according to claim 1, wherein a face region is detected by detecting a face pattern for the top candidate. A face authentication device for performing face authentication of a person moving toward a specific position,
An imaging unit that is provided between the moving person and a specific position and that captures an image including at least the face of the moving person;
Face detection means for detecting the face area of the person from the image captured by the imaging means;
Authentication means for determining whether or not the person is a pre-registered person by comparing the image of the face area detected by the face detecting means with pre-registered dictionary information,
The face detection means calculates inter-image differences for a plurality of images captured by the imaging means, detects a change area, performs optical flow extraction processing on the obtained change area, and obtains the obtained optical flow Grouping processing is performed so that optical flows that are close in position and close to the flow direction and size are in the same group. A range in which the first derivative value between the positions where the first derivative of the obtained peripheral feature value is between positive and negative values is close to “0” is extracted as a candidate for the top of the head. A face authentication apparatus for detecting a face region by detecting a face pattern for a top candidate. A face authentication device for performing face authentication of a person moving toward a specific position,
An imaging unit that is provided between the moving person and a specific position and that captures an image including at least the face of the moving person;
Face detection means for detecting the face area of the person from the image captured by the imaging means;
Face orientation determination means for determining whether or not the image of the face area detected by the face detection means is front-facing;
Authentication means for determining whether or not the person is a pre-registered person by comparing the image of the face area determined to be front-facing by the face orientation determining means with dictionary information registered in advance;
Display means provided in the vicinity of the specific position and displaying an enlarged image of the face area detected by the face detection means for the moving person;
It is detected that the image of the face area displayed by the display means is likely to deviate from the imaging visual field of the imaging means, and the end of the enlarged display image is highlighted on the side that is likely to deviate from the imaging visual field. Display control means for controlling the display means,
A face authentication apparatus comprising:   The face orientation determining means is configured so that when the value obtained by dividing the shorter length by the longer length by the height and width of the circumscribed rectangle of the face area detected by the face detecting means is equal to or greater than a threshold value, the face is facing forward If the value obtained by dividing the shorter length by the longer length is equal to or smaller than a threshold value, face pattern detection is performed on the face area, and face detection created in advance from a face facing front is performed. The similarity with the dictionary pattern is calculated, and when the obtained similarity is less than or equal to the threshold, it is determined that the face is not front-facing, and when the obtained similarity is greater than or equal to the threshold, the face is determined to be front-facing The face authentication apparatus according to claim 4, wherein: A face authentication device for performing face authentication of a person moving toward a specific position,
An imaging unit that is provided between the moving person and a specific position and that captures an image including at least the face of the moving person;
Face detection means for detecting the face area of the person from the image captured by the imaging means;
Face orientation determination means for determining whether or not the image of the face area detected by the face detection means is front-facing;
Authentication means for determining whether or not the person is a pre-registered person by collating an image of the face area determined to be front-facing by the face orientation determining means with dictionary information registered in advance. Configured,
The face orientation determining means is configured so that when the value obtained by dividing the shorter length by the longer length by the height and width of the circumscribed rectangle of the face area detected by the face detecting means is equal to or greater than a threshold value, the face is facing forward If the value obtained by dividing the shorter length by the longer length is equal to or smaller than a threshold value, face pattern detection is performed on the face area, and face detection created in advance from a face facing front is performed. The similarity with the dictionary pattern is calculated, and when the obtained similarity is less than or equal to the threshold, it is determined that the face is not front-facing, and when the obtained similarity is greater than or equal to the threshold, it is determined that the face is front-facing A face authentication apparatus characterized by that.   2. The apparatus according to claim 1, further comprising: a guide unit that guides the person to face the imaging unit or the display unit when the face direction determination unit determines that the face is not front-facing. Item 5. The face authentication device according to Item 4. A face authentication method for performing face authentication of a person moving toward a specific position,
An imaging step of imaging an image including at least a face of the moving person by an imaging unit provided between the moving person and a specific position;
A face detection step of detecting a face area of the person from the image captured by the imaging step;
An authentication step for determining whether or not the person is a pre-registered person by comparing the image of the face area detected by the face detection step with pre-registered dictionary information;
A display step of enlarging and displaying an image of the face area detected by the face detection step on the moving person by a display means provided in the vicinity of the specific position;
It is detected that the image of the face area displayed by this display step is likely to deviate from the imaging field of view of the imaging means, and the end of the enlarged display image is highlighted on the side that is likely to deviate from the imaging field of view. A display control step for controlling the display means;
A face authentication method comprising: A face authentication method for performing face authentication of a person moving toward a specific position,
An imaging step of imaging an image including at least a face of the moving person by an imaging unit provided between the moving person and a specific position;
A face detection step of detecting a face area of the person from the image captured by the imaging step;
An authentication step of determining whether the person is a pre-registered person by comparing the image of the face area detected by the face detection step with dictionary information registered in advance,
The face detection step calculates an inter-image difference for a plurality of images picked up by the image pickup step, detects a change region, performs an optical flow extraction process on the obtained change region, and obtains the obtained optical flow The grouping process is performed so that optical flows that are close in position and close to the flow direction and size are in the same group, and the peripheral characteristics for the groups in which the number of optical flows that make up the obtained group is greater than the threshold value. A range in which the first derivative value between the positions where the first derivative of the obtained peripheral feature value is between positive and negative values is close to “0” is extracted as a candidate for the top of the head. A face authentication method for detecting a face region by detecting a face pattern for a top candidate. A face authentication method for performing face authentication of a person moving toward a specific position,
An imaging step of imaging an image including at least a face of the moving person by an imaging unit provided between the moving person and a specific position;
A face detection step of detecting a face area of the person from the image captured by the imaging step;
A face orientation determination step of determining whether or not the image of the face area detected by the face detection step is front-facing;
An authentication step of determining whether or not the person is a pre-registered person by comparing the image of the face area determined to be front-facing by the face direction determination step with dictionary information registered in advance;
A display step of enlarging and displaying an image of the face area detected by the face detection step on the moving person by a display means provided in the vicinity of the specific position;
It is detected that the image of the face area displayed by this display step is likely to deviate from the imaging field of view of the imaging means, and the end of the enlarged display image is highlighted on the side that is likely to deviate from the imaging field of view. A display control step for controlling the display means;
A face authentication method comprising: A face authentication method for performing face authentication of a person moving toward a specific position,
An imaging step of imaging an image including at least a face of the moving person by an imaging unit provided between the moving person and a specific position;
A face detection step of detecting a face area of the person from the image captured by the imaging step;
A face orientation determination step of determining whether or not the image of the face area detected by the face detection step is front-facing;
An authentication step for determining whether or not the person is a pre-registered person by collating an image of the face area determined to be front-facing in the face direction determination step with dictionary information registered in advance. Configured,
In the face orientation determination step, the face is front-facing when a value obtained by dividing the shorter length by the longer length by the height and width of the circumscribed rectangle of the face area detected by the face detection step is equal to or greater than a threshold value. If the value obtained by dividing the shorter length by the longer length is equal to or smaller than a threshold value, face pattern detection is performed on the face area, and face detection created in advance from a face facing front is performed. The similarity with the dictionary pattern is calculated, and when the obtained similarity is less than or equal to the threshold, it is determined that the face is not front-facing, and when the obtained similarity is greater than or equal to the threshold, it is determined that the face is front-facing A face authentication method characterized by the above. An entrance / exit management device that performs face authentication of a person moving toward an entrance / exit gate provided in an entrance / exit area, and controls opening / closing of the entrance / exit gate based on the authentication result,
An imaging unit that is provided between the moving person and the entrance / exit gate, and that captures an image including at least the face of the moving person;
Face detection means for detecting the face area of the person from the image captured by the imaging means;
Authentication means for determining whether the person is a pre-registered person by collating an image of the face area detected by the face detection means with pre-registered dictionary information;
Gate control means for controlling opening / closing of the entrance / exit gate based on the determination result of the authentication means;
Display means provided in the vicinity of the entrance / exit gate and displaying an enlarged image of the face area detected by the face detection means for the moving person;
It is detected that the image of the face area displayed by the display means is likely to deviate from the imaging visual field of the imaging means, and the end of the enlarged display image is highlighted on the side that is likely to deviate from the imaging visual field. Display control means for controlling the display means,
The entrance / exit management apparatus characterized by comprising. An entrance / exit management device that performs face authentication of a person moving toward an entrance / exit gate provided in an entrance / exit area, and controls opening / closing of the entrance / exit gate based on the authentication result,
An imaging unit that is provided between the moving person and the entrance / exit gate, and that captures an image including at least the face of the moving person;
Face detection means for detecting the face area of the person from the image captured by the imaging means;
Authentication means for determining whether the person is a pre-registered person by collating an image of the face area detected by the face detection means with pre-registered dictionary information;
A gate control means for controlling the opening and closing of the entrance / exit gate based on the determination result of the authentication means,
The face detection means calculates inter-image differences for a plurality of images captured by the imaging means, detects a change area, performs optical flow extraction processing on the obtained change area, and obtains the obtained optical flow The grouping process is performed so that optical flows that are close in position and close to the flow direction and size are in the same group, and the peripheral characteristics for the groups in which the number of optical flows that make up the obtained group is greater than the threshold value. A range in which the first derivative value between the positions where the first derivative of the obtained peripheral feature value is between positive and negative values is close to “0” is extracted as a candidate for the top of the head. An entrance / exit management device for detecting a face area by detecting a face pattern of a candidate for the top of the head. An entrance / exit management device that performs face authentication of a person moving toward an entrance / exit gate provided in an entrance / exit area, and controls opening / closing of the entrance / exit gate based on the authentication result,
An imaging unit that is provided between the moving person and the entrance / exit gate, and that captures an image including at least the face of the moving person;
Face detection means for detecting the face area of the person from the image captured by the imaging means;
Face orientation determination means for determining whether or not the image of the face area detected by the face detection means is front-facing;
Authentication means for determining whether or not the person is a pre-registered person by comparing the image of the face area determined to be front-facing by the face orientation determining means with dictionary information registered in advance;
Gate control means for controlling opening / closing of the entrance / exit gate based on the determination result of the authentication means;
Display means provided in the vicinity of the entrance / exit gate and displaying an enlarged image of the face area detected by the face detection means for the moving person;
It is detected that the image of the face area displayed by the display means is likely to deviate from the imaging visual field of the imaging means, and the end of the enlarged display image is highlighted on the side that is likely to deviate from the imaging visual field. Display control means for controlling the display means,
The entrance / exit management apparatus characterized by comprising. An entrance / exit management device that performs face authentication of a person moving toward an entrance / exit gate provided in an entrance / exit area, and controls opening / closing of the entrance / exit gate based on the authentication result,
An imaging unit that is provided between the moving person and the entrance / exit gate, and that captures an image including at least the face of the moving person;
Face detection means for detecting the face area of the person from the image captured by the imaging means;
Face orientation determination means for determining whether or not the image of the face area detected by the face detection means is front-facing;
Authentication means for determining whether or not the person is a pre-registered person by comparing the image of the face area determined to be front-facing by the face orientation determining means with dictionary information registered in advance;
A gate control means for controlling the opening and closing of the entrance / exit gate based on the determination result of the authentication means,
The face orientation determining means is configured so that when the value obtained by dividing the shorter length by the longer length by the height and width of the circumscribed rectangle of the face area detected by the face detecting means is equal to or greater than a threshold value, the face is facing forward If the value obtained by dividing the shorter length by the longer length is equal to or smaller than a threshold value, face pattern detection is performed on the face area, and face detection created in advance from a face facing front is performed. The similarity with the dictionary pattern is calculated, and when the obtained similarity is less than or equal to the threshold, it is determined that the face is not front-facing, and when the obtained similarity is greater than or equal to the threshold, it is determined that the face is front-facing An entrance / exit management device characterized by that.

Images