JP2011192157A - Emergency report device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an emergency report device detecting an occurrence of abnormal situations and reporting the occurrence of the abnormal situations without being noticed by a burglar. <P>SOLUTION: The emergency report device 1 has: a collation part 63 for determining whether a person's image included in a monitor image that photographs a monitoring area is an image of a regular user pre-registered; an emergency action detection part 64 for determining whether the person is a regular user and whether the person takes prescribed emergency action on the basis of the monitoring image; and a report part 67 for reporting an emergency when the emergency action detection part 64 determines the person has taken the emergency action. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to an emergency call device that makes an emergency call when an abnormality is detected based on an image of a person in a monitoring area.

  As an emergency notification means when a bandit such as a robber enters the store, a button switch for emergency notification is provided near the place where the store clerk normally exists, such as near a cash register. However, in some cases, the store clerk may be forced to take a specific attitude so that the store clerk does not press the emergency button. Therefore, even in such a case, in order to be able to make an emergency call, a technique for making an emergency call when an abnormality is detected by analyzing an image of a monitoring area such as in a store has been proposed (for example, (See Patent Documents 1 and 2).

  For example, an active surveillance camera disclosed in Patent Document 1 extracts a human component from image data using a plurality of image frames captured by an imaging unit, extracts a head and both hands from the human component, Alternatively, it recognizes that the head is placed and recognizes that an abnormality has occurred, and notifies the outside of the occurrence of the abnormality when the occurrence of the abnormality is recognized.

  In addition, the monitoring system disclosed in Patent Document 2 identifies a person who is conscious of a plurality of cameras, monitors the behavior state of the identified person, and the identified person image is an image of a cheating addict, etc. It is determined whether or not the person performs cheating depending on whether or not the image matches a previously stored image and whether or not the action pattern of the specified person matches a specific reference pattern.

JP 2002-35459 A JP 2003-169320 A

However, in the technology disclosed in Patent Document 1, in the technique of detecting abnormalities by detecting conspicuous behavior such as a store clerk holding both hands overhead, a bandit prohibits the clerk from performing such conspicuous behavior. In such a case, the occurrence of an abnormal situation cannot be detected and the report will be lost.
In addition, since the technique disclosed in Patent Document 2 cannot prepare images of all suspicious persons in advance, it can be detected even if a person whose image is not stored in the device takes suspicious behavior or robbery Can not do it.

  Accordingly, an object of the present invention is to provide an emergency call device that can detect the occurrence of an abnormal situation and report the occurrence of the abnormal situation without being recognized by a bandit.

  As one aspect of the present invention, an emergency call device is provided. The emergency call device includes a collation unit that determines whether or not a person image included in a monitoring image obtained by photographing a monitoring area is an image of a regular user registered in advance, the person is a regular user, and monitoring The image includes an emergency action detection unit that determines whether or not the person has taken a predetermined emergency action in the image, and a notification unit that makes an emergency call when the emergency action detection unit determines that the person has taken an emergency action.

  In this emergency call device, when the collation unit determines that the person shown in the monitoring image is an unauthorized user who is not a regular user, whether or not the person has taken a predetermined suspicious behavior in the monitoring image It is preferable to further include a suspicious behavior detection unit for determining. In this case, when the reporting unit determines that the suspicious behavior detection unit has taken suspicious behavior, it is preferable to issue an alarm different from the emergency reporting.

  In addition, it is preferable that the emergency notification device further includes a face direction detection unit that detects the face direction of the person shown in the monitoring image. In this case, when the emergency direction detection unit detects the face direction of the person shown in the monitoring image in the first direction in the monitoring image taken in time series, the face direction detection unit detects the face direction of the person. It is preferable to determine that has taken emergency action.

  Alternatively, the suspicious behavior detection unit may detect a suspicious behavior when the face direction detection unit detects the face orientation of the person in the monitoring image in the second direction for a predetermined number of times in the monitoring images taken in time series. It is preferable to determine that the image has been taken.

  Furthermore, the face direction detection unit calculates a direct visibility representing the degree of the person facing the front face in the monitoring image from the face orientation, and the emergency action detection unit has the direct visibility equal to or greater than a predetermined threshold value. And face orientation in the first direction.

  The emergency notification device according to the present invention has an effect of detecting the occurrence of an abnormal situation and notifying the occurrence of the abnormal situation without being recognized by the bandits.

It is a schematic structure figure of an emergency call device concerning one embodiment of the present invention. It is a figure which shows an example of the coordinate system which prescribes | regulates the direction of a target person's face. (A) is a schematic diagram of a face image of the target person when the target person is facing the imaging unit, and (b) is a target when the target person is facing away from the imaging unit It is the schematic of a person's face image. (A) is a figure which shows the outline of the emergency action table which records the registration face image and emergency action counter for every regular user, (b) shows the registration face image and suspicious action counter for every non-regular user. It is a figure which shows the outline of the suspicious action table to record. It is a flowchart which shows the operation | movement of an emergency call process. It is a flowchart which shows the operation | movement of an emergency call process.

Hereinafter, an emergency call device according to an embodiment of the present invention will be described with reference to the drawings. This emergency call device determines whether a person shown in an image captured by a camera that captures a monitoring area such as a store is a registered regular user or a non-regular user who is not a regular user. And this emergency call device analyzes the image, detects the orientation of the face of the authorized user shown in the image, and whether the authorized user is taking a preset emergency action according to the face orientation. If an emergency action is determined, an emergency call is made.
When the emergency call device is installed in the store, the authorized user is, for example, the store clerk or owner of the store where the emergency call device 1 is installed, and the unauthorized user is, for example, a customer who visits the store. It is. When the emergency call device is installed in an office building, the authorized user is, for example, a person working in the office in the office building or an administrator of the office building, and the unauthorized user is, for example, an office building. A visitor to the office in the house.

  FIG. 1 is a diagram illustrating a schematic configuration of an emergency call device 1 according to an embodiment. As shown in FIG. 1, the emergency call device 1 includes an imaging unit 2, an interface unit 3, a storage unit 4, an output unit 5, and an abnormal behavior detection unit 6.

  The imaging unit 2 is a monitoring camera that captures a predetermined monitoring area, and is, for example, a photoelectric conversion element (for example, a CCD sensor, a C-MOS, or the like) that is arranged two-dimensionally and outputs an electrical signal corresponding to the amount of received light. And an imaging optical system for forming an image of the monitoring region on the photoelectric conversion element.

The imaging unit 2 is installed so as to photograph the direction of the flow line at a position along the flow line of the monitoring region so that the face of a person passing through or staying in the monitoring region can be photographed.
For example, when the emergency call device 1 monitors a visitor of a store such as a convenience store or a supermarket, the imaging unit 2 has a ceiling or a wall in the store so that a shooting area includes a passage near a cash register or between product shelves. It is attached in a state where the shooting direction is slightly downward. Furthermore, the imaging unit 2 may be attached so that the shooting range of the imaging unit 2 includes the entrance / exit. Moreover, when the emergency call apparatus 1 monitors an indoor passerby in an office building, the imaging unit 2 is attached, for example, above the ceiling or wall of a corridor with the shooting direction slightly downward. Further, the imaging unit 2 is not limited to one, and a plurality of imaging units 2 may exist.

  The imaging unit 2 acquires one frame of a monitoring image obtained by capturing the monitoring area at a predetermined time interval (for example, 200 msec). The imaging unit 2 is connected to the abnormal behavior detection unit 6 via the interface unit 3 and passes the acquired monitoring image to the abnormal behavior detection unit 6.

The interface unit 3 is an interface circuit connected to the imaging unit 2, and passes the monitoring image received from the imaging unit 2 to the abnormal behavior detection unit 6.
The storage unit 4 includes at least one of a semiconductor memory, a magnetic recording medium, its access device, an optical recording medium, and its access device. And the memory | storage part 4 memorize | stores the computer program for controlling the emergency call apparatus 1, various parameters, data, etc. In addition, the storage unit 4 stores an emergency action counter of a regular user registered in the emergency call apparatus 1 in advance, a suspicious action counter of a non-regular user registered in the emergency call apparatus 1, a face image, and the like. The details of the emergency action counter and the suspicious action counter will be described together with related components of the abnormal action detection unit 6.

  The output unit 5 is connected to the abnormal behavior detection unit 6 and has an interface circuit for connecting the emergency call device 1 to a communication line such as a public communication line or a local area network. Furthermore, the output unit 5 has an interface circuit for connecting to a monitor or speaker installed in a room in or near the monitoring area, or a wireless terminal carried by a regular user. When the abnormal action detection unit 6 detects a bandit such as a robber, the output unit 5 causes an abnormal situation to occur in a monitoring center device (not shown) installed in the monitoring center or police through the communication network. An emergency call signal indicating that Alternatively, when the abnormal behavior detection unit 6 detects a suspicious person in the monitoring area, the abnormal behavior detection unit 6 receives a warning voice signal or a warning message indicating that fact from the abnormal behavior detection unit 6 and outputs it to a speaker, a monitor, or a wireless terminal. .

The abnormal behavior detection unit 6 includes one or a plurality of processors and their peripheral circuits. Each time the abnormal behavior detection unit 6 receives a monitoring image from the imaging unit 2, the abnormal behavior detection unit 6 analyzes the behavior of the person shown in the monitoring image and needs an emergency call to the monitoring center device or a warning to a regular user. It is determined whether or not.
Therefore, the abnormal behavior detection unit 6 is a function module implemented by software that operates on the processor, and includes a person area detection unit 61, a face orientation detection unit 62, a face collation unit 63, an emergency behavior detection unit 64, a suspicious behavior, and the like. It has a detection unit 65, a suspicious person registration unit 66, and a notification unit 67.
Note that these units included in the abnormal behavior detection unit 6 may be configured by independent integrated circuits, firmware, a microprocessor, and the like.
Hereinafter, each part of the abnormal behavior detection unit 6 will be described in detail.

The person area detection unit 61 extracts a person area, which is an area in which a person is shown, from the monitoring image received from the imaging unit 2. For this purpose, the person area detection unit 61 calculates, for example, one or more feature amounts from a partial area on the monitoring image, and if the feature amount satisfies a predetermined condition considered to correspond to a person, the partial area Are extracted as a person area. Specifically, the person region detection unit 61 performs edge pixel extraction using a Sobel filter or the like so that the inclination direction of the luminance change can be known. The feature amount can be, for example, the direction distribution of the edge pixels obtained in the partial region, or the luminance distribution of pixels near the edge. In addition, the predetermined condition described above is satisfied by determining the range of the feature value such as the direction distribution of the edge pixel corresponding to the person and the luminance distribution near the edge from a plurality of images obtained by photographing the person. It can be determined in advance.
Alternatively, the person area detection unit 61 performs template matching with the monitoring image using a predetermined template, and extracts an area where the degree of matching obtained as a result of the template matching is equal to or greater than a predetermined value as a person area. You may make it do. In addition, the person area detection unit 61 performs a difference between the monitoring image and a reference image captured in a state where no moving object such as a person exists in the monitoring area, and pixels whose difference value is equal to or greater than a predetermined value are connected. Extracted difference area. And the person area detection part 61 may extract the difference area as a person area, when the difference area has a size and shape considered to correspond to a person. Various methods are already known as the person region extraction processing, and the description thereof is omitted here.

  When extracting the person area, the person area detecting unit 61 extracts an area corresponding to the head from the person area. The person area detection unit 61 creates, for example, an edge image obtained by extracting an edge from the person area. The person region detection unit 61 can detect an elliptical edge distribution that approximates the contour shape of the head from the edge image, and extract a region surrounded by the edge distribution as a head region. In this case, the person region detection unit 61 can detect an elliptical edge distribution using, for example, a generalized Hough transform. Alternatively, the person area detection unit 61 performs template matching on the person area or the edge image of the person area using the edge direction distribution, the luminance distribution, and the like as feature amounts using a head template learned in advance, An area where the degree of matching obtained as a result of the template matching is equal to or greater than a predetermined value can be extracted as a head area. Furthermore, the person area detection unit 61 may use an area considered to correspond to the head of the person area as the head area (for example, from the upper end of the person area, corresponding to the average head size and shape). Cut out the area and use it as the head area). Furthermore, the person region detection unit 61 may extract the head region using another head region extraction method. For example, the human region detection unit 61 may use a face detector disclosed in “Rapid Object Detection using a Boosted Cascade of Simple Features”, Proceedings IEEE Conf. On Computer Vision and Pattern Recognition, 2001.

  The person area detection unit 61 creates image data of the head area obtained by cutting out the head area from the monitoring image. Then, the person area detecting unit 61 assigns a person identifier to be used for identification with other head areas to the image data of the head area, and uses them as one person information. The person area detection unit 61 stores the obtained person information in the storage unit 4 so that it can be used by other units of the abnormal behavior detection unit 6. In addition, when there are a plurality of head regions in one monitoring image, the person region detection unit 61 extracts all head regions, creates personal information, and stores them in the storage unit 4. At this time, the person identifier is set to a different value for each head region.

The face orientation detection unit 62 detects the face orientation based on the front orientation of the person shown in the head region.
FIG. 2 shows an orthonormal coordinate system for defining the face orientation of a person. In this orthonormal coordinate system (X, Y, Z), the origin O is set to the approximate center of the head of the person 200 whose face orientation is to be estimated. The Z axis is a straight line connecting the head of the person 200 and the imaging unit 2. The X axis passes through the origin O and is set to a vertical straight line orthogonal to the Z axis. Similarly, the Y axis passes through the origin O and is set to a horizontal straight line orthogonal to the Z axis.
In this coordinate system, the face direction of the person 200 corresponds to a state in which the person 200 is directly facing the imaging unit 2 (that is, the person 200 is facing the front of the face to the imaging unit 2). It is represented by a set (ψ, θ) of a yaw angle ψ and a pitch angle θ. The yaw angle ψ represents a rotation angle in the left-right direction with respect to the face direction of the person 200 in the face-to-face state (that is, a rotation angle in the YZ plane with respect to the Z axis). Further, the pitch angle θ represents a vertical rotation angle (that is, a rotation angle in the XZ plane with respect to the Z axis) with respect to the face direction of the person 200 in the face-to-face state. In the following, the yaw angle ψ and the pitch angle θ are expressed in radians, and the rightward direction or the downward direction is positive, respectively.

  The face orientation detection unit 62 reads out image data of the head region from the storage unit 4 and extracts facial feature points that are characteristic portions of the face from the head region. Then, the face orientation detection unit 62 calculates the type of the extracted facial feature point and position information on the head region (for example, a two-dimensional coordinate value with the upper left corner of the head region as the origin). For example, the face orientation detection unit 62 extracts facial feature points such as both eye corners, both eye region centers, nose tips, mouth points, and mouth corner points. The face direction detection unit 62 can use various known methods for extracting face feature points. For example, the face orientation detection unit 62 performs edge extraction processing on the head region to extract edge pixels having a large luminance difference from surrounding pixels. Note that such edge pixels in the head region may be extracted from the edge image of the person image. Then, the face direction detection unit 62 checks whether or not the feature amount obtained based on the position and pattern of the edge pixel satisfies a predetermined condition for a characteristic part of the face such as eyes, nose, and mouth. By specifying the position of each part, each facial feature point can be extracted. In addition, instead of performing edge extraction processing to extract edge pixels, the face orientation detection unit 62 performs Gabor conversion processing or wavelet conversion processing to extract pixels that have large local changes in different spatial frequency bands. May be. Further, the face direction detection unit 62 may extract face feature points by performing template matching between a template corresponding to each part of the face and a head region to identify the position of each part of the face.

When the face feature points are extracted, the face direction detection unit 62 examines the two-dimensional distribution of each face feature point and detects the face direction. That is, in the present embodiment, the face direction can be said to be the amount of deviation of the feature points from the face facing front (facing state).
The relationship between the face orientation and the face feature points will be described with reference to FIG. FIG. 3A shows a face image 310 of the target person when the target person is facing the image pickup unit 2 (that is, when the target person is in a directly-facing state). On the other hand, FIG. 3B shows a face image 330 of the target person when the target person is facing away from the imaging unit 2. As illustrated in FIG. 3A, when the target person faces the front of the face with respect to the imaging unit 2, the nose tip 311 that is the apex of the nose exists on the midline 320 of the face. Also, the left and right eye corners 312 and 313, the left and right mouth corner points 314 and 315, and the like can be observed.

  On the other hand, as shown in FIG. 3B, when the target person is facing away from the imaging unit 2, the nose tip 331 protrudes from the eyes and mouth, so that the yaw angle ψ increases. , Away from the midline 340. Further, the left and right eye corners 332 and 333 and the left and right mouth corner points 334 and 335 are hidden and become invisible as the target person rotates the face. For example, when the target person rotates his face a little more to the right than shown in FIG. 3B, the right eye corner 333 or the right mouth corner point 335 is not reflected on the face image 330, and the face feature points cannot be detected. It becomes possible. Thus, the orientation of the face of the target person relative to the imaging unit 2 can be examined from the position of the nose tip with respect to the midline of the face and the type of face feature points that cannot be detected.

  Therefore, the face direction detection unit 62 obtains, for example, a straight line connecting the midpoint of the eyes and the mouth point as the midline of the face. Then, the face direction detection unit 62 determines whether the nose apex, which is the apex of the nose, is on the left or right side of the midline and the distance from the midline. In addition, the face orientation detection unit 62 examines in advance the relationship between the direction of the median line and the distance from the median line to the nose tip as information on the facial feature point arrangement, and the orientation of the face including at least the front face. Is stored in the storage unit 4 in advance. Then, the face direction detection unit 62 refers to the look-up table to determine the face direction corresponding to the obtained midline direction and the distance from the midline to the nose tip.

Further, the face orientation detection unit 62 may estimate the face orientation based on the positional relationship between the extracted face feature points and the corresponding face feature points in the three-dimensional shape model representing the three-dimensional shape of the human face. . In this case, a standard three-dimensional shape model imitating a human head (for example, a wire frame model or a surface model) is prepared in advance and stored in the storage unit 4. In addition, the position of 3D face feature points such as the corners of both eyes, the tip of the nose and the mouth point is also stored in the storage unit 4 as information on the arrangement of the face feature points of the three-dimensional shape model.
When face feature points are extracted from the head region, the face orientation detection unit 62 adjusts the face orientation of the three-dimensional shape model according to a predetermined rotation amount, translation amount, or enlargement / reduction ratio, and the imaging unit 2 Is virtually projected onto a plane parallel to the image plane of the imaging optical system, and the position of the 3D face feature point on the plane is obtained. Then, the face orientation detection unit 62 obtains the sum of the positional deviation amounts between each face feature point extracted from the head region and the corresponding feature point among the projected 3D face feature points. The face orientation detection unit 62 changes the rotation amount, the translation amount, or the enlargement / reduction ratio, repeats the above procedure, and obtains the face orientation of the three-dimensional shape model when the sum of the positional deviation amounts is minimized. Then, the face orientation detection unit 62 can obtain the above-mentioned set (ψ, θ) of the yaw angle ψ and the pitch angle θ from the face orientation of the three-dimensional shape model. In this case as well, the face orientation is calculated from the positional relationship of the feature points with reference to the front orientation, that is, it can be said that the amount of deviation of the feature points from the face facing the front.
Note that the face direction detection unit 62 projects each face feature point extracted from the head region into the three-dimensional space, so that the amount of positional deviation from the 3D face feature point on the three-dimensional shape model is minimized. The orientation of the face of the three-dimensional shape model may be determined by performing processing such as rotation and enlargement / reduction on the three-dimensional shape model.
When the face direction detection unit 62 obtains the values of the yaw angle ψ and the pitch angle θ representing the face direction, it stores them in the storage unit 4 in association with the corresponding person information as the face direction information.

ここで、角度ψは、図２に示されるヨー角であり、頭部の中心を原点として、頭部領域に写っている人物がカメラを直視した状態、すなわち、その人物の顔の正面を撮像部２に対向させている状態を正対状態（全ての回転角が０）とした場合の左右方向（Ｘ軸回り）の回転角である。また角度θは、ピッチ角であり、正対状態におけるその人物の顔向きに対する、上下方向（Ｙ軸回り）の回転角である。
なお、直視度LSが無限大に近い値となることを防止するために、ヨー角ψ及びピッチ角θが所定の下限値（例えば、0.1ラジアン）以下となる場合には、非常行動検出部６４は、ヨー角ψ及びピッチ角θをその下限値に設定して、直視度LSを求めてもよい。 Furthermore, the face direction detection unit 62 is such that the person in the head region has the face facing the image pickup unit 2, that is, the face is directed in the direction in which the face appears in the front direction in the monitoring image. As an index representing, the camera direct visibility is calculated based on the face orientation information. This camera direct view degree LS is used for detecting an emergency action and a suspicious action, which will be described later, and is calculated according to the following equation.

Here, the angle ψ is the yaw angle shown in FIG. 2, and the person in the head region looks directly at the camera with the center of the head as the origin, that is, the front of the person's face is imaged. This is the rotation angle in the left-right direction (around the X axis) when the state facing the part 2 is the directly-facing state (all rotation angles are 0). Further, the angle θ is a pitch angle, and is a rotation angle in the vertical direction (around the Y axis) with respect to the face direction of the person in the face-to-face state.
In order to prevent the direct visibility LS from becoming a value close to infinity, when the yaw angle ψ and the pitch angle θ are equal to or less than a predetermined lower limit (for example, 0.1 radians), the emergency action detection unit 64 May determine the direct visibility LS by setting the yaw angle ψ and the pitch angle θ to their lower limits.

The direct visibility LS and the orientation of the target person's face will be described with reference to FIGS. As shown in FIG. 3A, when the target person is substantially facing the front with respect to the imaging unit 2, both the yaw angle ψ and the pitch angle θ are small values, and the direct visibility LS is a large value. It becomes. On the other hand, as shown in FIG. 3B, when the target person tilts his / her face with respect to the imaging unit 2, at least one of the yaw angle ψ and the pitch angle θ has a large value. Is a relatively small value.
The face orientation detection unit 62 stores the calculated camera direct visibility in the storage unit 4 in association with the corresponding person information.

  The face collation unit 63 uses the image data of the head region extracted from the monitoring image and the registered user's registered facial image stored in the storage unit 4 or a non-regular use registered as a possible suspicious person. Against the person's face image. Then, the face matching unit 63 determines whether or not the person shown in the head region is authenticated as a regular user registered in the emergency call device 1. Or the face collation part 63 determines whether the person reflected in the head area | region corresponds to any of the unauthorized users registered.

The face matching unit 63 can use any of various known matching methods in order to match the face shown in the head region extracted from the monitoring image and the face shown in the face image. For example, the face matching unit 63 performs pattern matching between the head region and the face images of each regular user and registered non-regular users. Then, the face matching unit 63 performs normalization by dividing the sum of squares of the luminance values of the pixels in the face area included in the head area and the corresponding pixels of the face image by the number of pixels included in the face area of the head area. The reciprocal of the converted value is obtained as the degree of coincidence. The face collation unit 63 selects a person in the head region from among the matching degrees obtained for each face image when the highest matching score exceeds a predetermined reference value, and the face image having the highest matching score. It is determined that it corresponds to a regular user corresponding to or a registered non-regular user. On the other hand, if the matching level does not exceed a predetermined reference value, the face matching unit 63 is not a regular user who is in the head region and a registered non-regular user. Judge that there is no.
When the face collating unit 63 determines that the person shown in the head area is any regular user, the face collating unit 63 associates the identification code of the regular user with the person information including the image data of the head area. Similarly, if the face collation unit 63 determines that the person shown in the head area is one of the registered non-regular users, the human face information including the image data of the head area is displayed as non-human information. Associate a legitimate user identification code. Further, when the face matching unit 63 determines that the person in the head area is neither a regular user nor an unauthorized user, the face matching unit 63 stores unregistered person information including image data of the head area. Associate a flag indicating that it is an unauthorized user.

Whether the emergency action detection unit 64 is taking an emergency action based on the orientation of the face of the person on the plurality of continuously captured images taken for the person determined to be a regular user by the face matching unit 63 Determine whether.
Here, the emergency action is the action of a regular user taken in an abnormal situation that requires reporting to the police or the monitoring center of the security company without any delay, such as the occurrence of a burglary. This emergency action is an action designated in advance by the emergency call device 1. In this embodiment, as an example of the emergency action, a regular user directly views the imaging unit 2 over a certain period.

The emergency action detection unit 64 stores, in the storage unit 4, the direct-view degree LS associated with the person information created for the monitoring image, which is associated with the authorized user identification code, every time the monitoring image is acquired. And the direct visibility LS is determined to be greater than or equal to a predetermined threshold. For example, the predetermined threshold is set to 4 corresponding to about 15 ° from the direction facing the imaging unit 2. If the degree of direct vision is equal to or greater than a predetermined threshold, there is a possibility that the legitimate user faces directly toward the imaging unit 2 and is directly looking and taking an emergency action. Therefore, the emergency action detection unit 64 uses the emergency action counter stored in the storage unit 4 and corresponding to the authorized user to determine whether the authorized user has directly viewed the imaging unit 2 for a certain period. To do.
The emergency action counter is a value that represents the number of times that an authorized user has performed an action corresponding to an emergency action in units of the number of monitoring images, and the initial value is set to zero. The emergency action detection unit 64 increments the value of the emergency action counter stored in the storage unit 4 and corresponding to the normal user if the direct visibility is equal to or greater than a predetermined threshold. On the other hand, when the authorized user turns away from the imaging unit 2, the emergency action detection unit 64 resets the value of the emergency action counter for the authorized user to 0.

  FIG. 4A is a diagram illustrating an outline of an emergency action table that stores a registered face image and an emergency action counter for each authorized user, which are stored in the storage unit 4. Each column of the leftmost column 401 of the emergency action table 400 stores the identification code of the authorized user. Each column of the central column 402 of the emergency action table 400 stores data representing a registered user's face image corresponding to the identification code shown in the leftmost column of the same row. Note that the registered face image itself may be stored separately from the emergency action table 400, and each column in the center column 402 of the emergency action table 400 may store an address number representing a recording area of the registered face image. Each column of the rightmost column 403 of the emergency behavior table 400 stores an emergency behavior counter for a regular user corresponding to the identification code indicated in the leftmost column of the same row.

  The emergency action detection unit 64 determines that an emergency action has been taken when it is determined that the authorized user is looking directly at the imaging unit 2 for a certain period of time. Therefore, the emergency action detection unit 64 compares the value of the emergency action counter with the emergency action determination threshold, and determines that the emergency action is taken when the value of the emergency action counter is equal to or greater than the emergency action determination threshold. Then, the emergency action detection unit 64 returns a determination result indicating that the emergency action has been detected to the abnormal action detection unit 6. Note that the emergency action determination threshold is set according to the shooting interval of the imaging unit 2, for example. For example, the emergency action determination threshold is set to 15 when the shooting interval of the imaging unit 2 is set to 200 milliseconds and the above-described fixed period is 3 seconds.

Whether the suspicious behavior detection unit 65 is performing suspicious behavior on the person determined to be an unauthorized user by the face matching unit 63 based on the orientation of the person's face on a plurality of continuously captured monitoring images. Determine whether or not.
Here, the suspicious behavior is an action of an unauthorized user, unlike an emergency action, and it is not necessary to immediately report to the police or the like, but an action that may soon lead to a criminal act. As an example, in order to confirm where the imaging unit 2 is installed in the monitoring area, a behavior in which an unauthorized user repeats a glance at the imaging unit 2 is regarded as a suspicious behavior. For example, when a non-regular user attempts shoplifting, such an action may be performed when confirming where the imaging unit 2 is installed in a store that is a monitoring area so as not to detect shoplifting, or in the near future. This is done when pre-viewing the inside of the store in advance in order to conduct illegal activities such as robbery.

  The suspicious behavior detection unit 65 determines whether or not the unauthorized user is directly looking at the imaging unit 2 based on the face orientation of the unauthorized user on the monitoring image. Therefore, each time a monitoring image is acquired, the suspicious behavior detection unit 65 includes a registered non-regular user identification code or non-regularity in the direct-view degree LS associated with the personal information created for the monitoring image. What is associated with the flag representing the authorized user is read from the storage unit 4, and it is determined whether or not the degree of direct view is equal to or greater than a predetermined threshold. The predetermined threshold is set to 4, for example, similar to the threshold used as a reference for determining that the authorized user is directly looking at the imaging unit 2. If the degree of direct vision is equal to or greater than a predetermined threshold, the non-regular user may be facing directly toward the imaging unit 2 and looking directly, and may be suspicious. Therefore, the suspicious behavior detection unit 65 uses the suspicious behavior counter stored in the storage unit 4 and corresponding to the unauthorized user to determine whether or not the authorized user has repeatedly glanced at the imaging unit 2. .

  The suspicious behavior counter is a value representing the number of times that an unauthorized user performs an action corresponding to suspicious behavior in units of the number of monitoring images. In addition, the non-regular user is not originally registered in the emergency call apparatus 1 like the customer who visited the store, for example. Therefore, it is determined by the face matching unit 63 that the person shown in the monitoring image is not a regular user and is not a registered non-regular user, and the suspicious behavior detection unit 65 has a direct visibility for the person. When it is determined that the predetermined threshold value is exceeded, the person is registered as an unauthorized user by the suspicious person registration unit 66 described later. Then, the registered suspicious behavior counter of the unauthorized user is set to 1 as an initial value.

  In addition, when the face matching unit 63 determines that the person shown in the monitoring image is a registered non-regular user, the suspicious behavior detection unit 65 stores the memory if the directivity is equal to or greater than a predetermined threshold. The value of the suspicious behavior counter corresponding to the registered irregular user stored in the unit 4 is incremented by one.

  FIG. 4B is a diagram showing an outline of a suspicious behavior table that stores a registered face image and a suspicious behavior counter for each non-regular user stored in the storage unit 4. In each column of the leftmost column 411 of the suspicious behavior table 410, an identification code of an unauthorized user is stored. Further, in each column of the central column 412 of the suspicious behavior table 410, data representing the registered face image of the unauthorized user corresponding to the identification code shown in the leftmost column of the same row is stored. Note that the registered face image itself is stored separately from the suspicious behavior table 410, and each column in the center column 412 of the suspicious behavior table 410 stores an address number representing the recording area of the registered face image of the unauthorized user. May be. Further, in each column of the rightmost column 413 of the suspicious behavior table 410, a suspicious behavior counter for an unauthorized user corresponding to the identification code indicated in the leftmost column of the same row is stored.

  The suspicious behavior detection unit 65 determines that the suspicious behavior is taken when the unauthorized user determines that the direct viewing of the imaging unit 2 is repeated. Therefore, the suspicious behavior detection unit 65 compares the value of the suspicious behavior counter with the suspicious behavior determination threshold, and determines that the suspicious behavior is taken when the value of the suspicious behavior counter becomes equal to or greater than the suspicious behavior determination threshold. Then, the suspicious behavior detection unit 65 returns a determination result indicating that the suspicious behavior is detected to the abnormal behavior detection unit 6. The suspicious behavior determination threshold is set to 15, for example.

  The suspicious person registration unit 66 determines that the person shown in the monitoring image is not a regular user and is not any registered unauthorized user, and the suspicious behavior detection unit 63 If 65 determines that the person is taking an action that may result in suspicious behavior (in this embodiment, if the person determines that the person has directly viewed the imaging unit 2), then the person is registered. Therefore, the suspicious person registration unit 66 generates an identification code different from the identification codes of the non-regular user and the regular user already registered as the identification code of the person newly registered as the non-regular user. . Then, the suspicious person registration unit 66 sets the head region of the person extracted from the monitoring image by the face direction detection unit 62 as a registered face image for the person. The suspicious person registration unit 66 stores the newly generated identification code and the registered face image in the suspicious action table. Furthermore, the suspicious person registration unit 66 sets the value of the suspicious behavior counter corresponding to the newly generated identification code to 1, and stores the value of the suspicious behavior counter in the suspicious behavior table.

When the emergency behavior detecting unit 64 detects that the authorized user has taken emergency behavior, the reporting unit 67 has detected that an abnormal situation has occurred in the monitoring center device (not shown) via the output unit 5. Make an emergency call.
Further, the reporting unit 67 issues a warning when the suspicious behavior detecting unit 65 detects that the unauthorized user has taken the suspicious behavior. For example, the reporting unit 67 outputs a warning voice or a message notifying that a suspicious person is in the monitoring area to a speaker or a monitor (not shown) via the output unit 5, or is possessed by a regular user. Ring the wireless terminal to be used. Further, the warning voice or the warning message can be, for example, a specific voice or message that only a regular user can know.

FIG. 5 and FIG. 6 are flowcharts showing the operation of the emergency call process controlled by the abnormal behavior detection unit 6.
First, the abnormal behavior detection unit 6 acquires a monitoring image from the imaging unit 2 via the interface unit 3 (step S101). Then, the abnormal behavior detection unit 6 passes the monitoring image to the person area detection unit 61 of the abnormal behavior detection unit 6.
The person area detection unit 61 extracts a person area in which a person is shown on the monitoring image, and further extracts a head area corresponding to the person's head from the person area (step S102). Then, the person area detection unit 61 determines whether or not a head area exists on the monitoring image (step S103). If the head region does not exist on the monitoring image, the abnormal behavior detection unit 6 repeats the processes of steps S101 to S103. On the other hand, when the head region exists on the monitoring image, the person region detection unit 61 cuts out the head region from the monitoring image, creates person information including image data representing the head region, and stores the storage unit 4. To remember.

  Next, the face orientation detection unit 62 of the abnormal behavior detection unit 6 reads the image data of the head region stored in the storage unit 4 and detects the orientation of the face of the person in the head region ( Step S104). Further, the face orientation detection unit 62 calculates the camera direct visibility based on the face orientation (step S105). The face orientation detection unit 62 stores the camera direct view degree in the storage unit 4 in association with the corresponding person information.

Further, the face matching unit 63 of the abnormal behavior detecting unit 6 reads the image data of the head region from the storage unit 4, and the face reflected in the head region and the regular user registered in the storage unit 4. By comparing with the registered face image or the registered face image of the registered non-regular user, it is determined whether or not the person shown in the head region is a regular user (step S106).
If the person is a regular user, the emergency behavior detection unit 64 of the abnormal behavior detection unit 6 determines that the regular user uses the imaging unit 2 based on the determination result of whether or not the camera direct visibility is equal to or greater than a predetermined threshold. It is determined whether the user is looking directly (step S107). If the authorized user is looking directly at the imaging unit 2, the emergency action detecting unit 64 increments the value of the emergency action counter corresponding to the authorized user by 1 (step S108). On the other hand, if the authorized user is not looking directly at the imaging unit 2, the emergency action detecting unit 64 resets the value of the emergency action counter corresponding to the authorized user to 0 (step S109).
After step S108 or S109, the emergency action detection unit 64 determines whether or not the value of the emergency action counter is equal to or greater than the emergency action threshold (step S110). If the value of the emergency action counter is less than the emergency action threshold, the abnormal action detection unit 6 repeats the processing from step S101.
On the other hand, if the value of the emergency action counter is equal to or greater than the emergency action threshold, the emergency action detection unit 64 determines that the regular user is taking an emergency action. Then, the reporting unit 67 of the abnormal behavior detecting unit 6 makes an emergency report to the monitoring center device via the output unit 5 (step S111). And the abnormal action detection part 6 repeats the process after step S101.

On the other hand, when it is determined in step S106 that the person is an unauthorized user (step S106-No), as shown in FIG. 6, the suspicious behavior detection unit 65 of the abnormal behavior detection unit 6 directly looks at the camera. Based on the determination result of whether or not the degree is equal to or greater than a predetermined threshold, it is determined whether or not the unauthorized user is directly looking at the imaging unit 2 (step S112). If the unauthorized user is looking directly at the imaging unit 2, the suspicious behavior detecting unit 65 determines whether the unauthorized user has already been registered based on the determination result of the face matching unit 63 ( Step S113). If already registered, the suspicious behavior detection unit 65 increments the value of the suspicious behavior counter corresponding to the unauthorized user by 1 (step S114). On the other hand, if the irregular user is not registered, the suspicious person registration unit 66 newly registers the irregular user. Specifically, the suspicious person registration unit 66 stores the image data of the head area of the unauthorized user in the storage unit 4 as the registered facial image of the unauthorized user and newly registered unauthorized use. The suspicious behavior counter of the person is set to 1 (step S115).
After step S114 or S115, the suspicious behavior detection unit 65 determines whether or not the value of the suspicious behavior counter is greater than or equal to the suspicious behavior threshold (step S116). If the value of the suspicious behavior counter is less than the suspicious behavior threshold value, the abnormal behavior detecting unit 6 repeats the processing after step S101.
On the other hand, if the value of the suspicious behavior counter is greater than or equal to the suspicious behavior threshold, the suspicious behavior detection unit 65 determines that the unauthorized user is taking suspicious behavior. Then, the reporting unit 67 of the abnormal behavior detecting unit 6 transmits a warning message to the wireless terminal possessed by the speaker, the monitor, or the authorized user via the output unit 5 (step S117).

After step S117 or in step S112, when it is determined that the non-regular user appearing in the head region is not looking directly at the imaging unit 2, the abnormal behavior detection unit 6 repeats the processing after step S101. .
If the abnormal behavior detection unit 6 detects a plurality of head regions from one monitoring image in step S102, the abnormal behavior detection unit 6 performs the processing from step S104 onward for each head region.

In addition, when the abnormal behavior detection unit 6 detects an emergency behavior and makes an emergency call for each of a plurality of (for example, any of 2 to 10) monitoring images acquired continuously, The emergency action counter may be reset to 0, and thereafter the emergency action counter may not be added until a predetermined time (for example, 10 minutes) elapses. Alternatively, the abnormal behavior detection unit 6 may reset each emergency user's emergency behavior counter to 0 after a predetermined time (for example, 1 minute) has elapsed since the first emergency notification.
In addition, when the abnormal behavior detection unit 6 once warns that a suspicious person has been detected, the abnormal behavior detection unit 6 may delete the identification code, the face image, and the suspicious behavior counter of the illegitimate user identified as the suspicious person from the suspicious behavior table. . Alternatively, the abnormal behavior detection unit 6 resets the suspicious behavior counter of the non-regular user determined as the suspicious user to 0, and thereafter, the suspicious behavior counter is incremented until a predetermined time (for example, 10 minutes) elapses. It may not be performed.

In addition, the abnormal behavior detection unit 6 detects the non-regular user identified as a suspicious user after a certain time (for example, one minute) has passed since the last warning, and the identification code and face image of the non-regular user. The suspicious behavior counter may be deleted from the suspicious behavior table.
Further, the suspicious behavior detection unit 65 may prevent a registered non-regular user from erroneously detecting, as suspicious behavior, the behavior of an unauthorized user who has just looked at the imaging unit 2 several times by chance. If the face is turned away from the imaging unit 2 over a predetermined release period, the registration of the unauthorized user may be deleted. In this case, the suspicious behavior detection unit 65 also stores the first time at which it is determined that the unauthorized user turns away from the imaging unit 2 in the storage unit 4 in association with the identification code of the unauthorized user. Note that the suspicious behavior detection unit 65 determines that the non-regular user faces the face from the imaging unit 2 if the directivity is less than the above-described threshold value used for determining that the non-regular user is directly viewing the imaging unit 2. Judging that Moreover, the cancellation | release period can be made into 3 minutes, for example.

  As described above, the emergency call device determines the orientation of the face of the authorized user extracted from the monitoring image, and based on the orientation of the face, the authorized user directly views the imaging unit for a certain period of time. When an emergency action designated in advance is detected, an emergency call is made. In this way, this emergency call device can detect that an abnormal situation has occurred based on the behavior of the authorized user and perform an emergency call even if the authorized user does not take noticeable action. Without being realized, it is possible to report that an abnormal situation has occurred. In addition, since this emergency call device determines whether or not to make an emergency call based on the action of the person who is determined to be an authorized user based on the monitoring image, a person other than the authorized user can behave like an emergency action. Even if you take it, you can prevent accidental emergency calls. Furthermore, since this emergency call device can detect that the unauthorized user has taken suspicious behavior based on the orientation of the unauthorized user's face in the surveillance image, there is a possibility that fraudulent activity may be performed. A certain user can be warned beforehand.

  In addition, this invention is not limited to said embodiment. For example, the storage unit may store a three-dimensional face model representing the three-dimensional shape of the regular user's face instead of storing the regular user's two-dimensional face image. In this case, the face matching unit projects each three-dimensional face model onto a two-dimensional plane so as to have the same orientation as the face of the person extracted from the monitoring image, and creates a two-dimensional face image. The projected two-dimensional face image is collated with the face image on the monitoring image.

The emergency action detection unit does not determine whether or not to increment the emergency action counter every time a monitoring image is acquired, but within a predetermined unit time every predetermined unit time (for example, 0.5 seconds or 1 second). If an emergency action is detected even once, an emergency action counter may be incremented.
In this case, the emergency action detecting unit is taking an emergency action if the period during which any authorized user is directly looking at the imaging unit is longer than a predetermined period (for example, 3 seconds). May be determined.
Similarly, the suspicious behavior detection unit may increment the suspicious behavior counter when the suspicious behavior is detected even once within the predetermined unit time every predetermined unit time. The suspicious behavior detection unit also has a suspicious behavior if any non-regular user is viewing the imaging unit directly for a longer period than a predetermined period (for example, 3 seconds). May be determined.

ここで、角度ψは、図２に示されるヨー角であり、正対状態（全ての回転角が０）における顔の向きに対する左右方向（Ｘ軸回り）の回転角である。また角度θは、ピッチ角であり、正対状態における顔向きに対する、上下方向（Ｙ軸回り）の回転角である。 Further, the suspicious behavior detection unit may detect the suspicious behavior of the unauthorized user based on behaviors other than the behavior of repeatedly glancing at the imaging unit. For example, the suspicious behavior detection unit may detect a behavior in which an unauthorized user intentionally turns away from the imaging unit for a certain period of time as a suspicious behavior. In this case, when it is determined that the person shown in the latest monitoring image is an unauthorized user, the suspicious behavior detecting unit represents the degree to which the unauthorized user is facing away from the imaging unit. The camera distance C _avoid is calculated. The camera distance C _avoid is calculated by the following equation, for example.

Here, the angle ψ is the yaw angle shown in FIG. 2, and is the rotation angle in the left-right direction (around the X axis) with respect to the face orientation in the face-to-face state (all rotation angles are 0). Further, the angle θ is a pitch angle, and is a rotation angle in the vertical direction (around the Y axis) with respect to the face direction in the directly-facing state.

The suspicious behavior detection unit determines that the unauthorized user is facing away from the imaging unit when the camera telescope C _{avoid of the unauthorized} user is equal to or greater than a predetermined threshold, and the suspicious behavior against the unauthorized user is determined. Increment the action counter by one. Note that the predetermined threshold is a value for determining that the face direction is more than a predetermined amount from a state in which the face direction is facing the imaging unit, and that the face is turned away from the imaging unit. It is a judgment standard for judging. For example, the threshold value may be a value corresponding to the face orientation when either the eyes corners or the corner points of the eyes of the viewer are no longer visible on the monitoring image. Alternatively, the boundary value obtained using a discriminant analysis or the like by photographing a person with the camera deliberately facing away from the camera in advance and obtaining a camera distance C _avoid in each state. May be set as a threshold value.
The suspicious behavior detection unit determines that the illegitimate user has taken suspicious behavior when the suspicious behavior counter for the specific unauthorized user becomes equal to or greater than the suspicious behavior threshold. The suspicious behavior threshold is a value corresponding to a non-regular user continuously facing away from the imaging unit over a certain period of time, and is, for example, the number of monitoring images captured in 5 minutes. be able to.

ここで、角度ψは、図２に示されるヨー角であり、正対状態（全ての回転角が０）における顔の向きに対する左右方向（Ｘ軸回り）の回転角である。また角度θは、ピッチ角であり、正対状態における顔向きに対する、上下方向（Ｙ軸回り）の回転角である。また角度ψ₀及びθ₀は、それぞれ、正規利用者が特定の場所がある方向を向いたときのヨー角及びピッチ角である。この特定の場所がある方向を表す角度ψ₀及びθ₀は、正規利用者がいる位置に応じて異なる値となる。一方、非常行動検出部は、監視画像上の正規利用者が写っている位置で、監視領域内に正規利用者の位置を特定できる。そこで記憶部に、その角度ψ₀及びθ₀を、予め監視画像上の各画素と対応付けて記憶させておく。そして非常行動検出部は、正規利用者が写っていると判断された頭部領域の重心が位置する画素に対応する角度ψ₀及びθ₀を記憶部から読み込んで、顔向き検出部により検出された正規利用者の顔の向きを表すヨー角ψ、ピッチ角θとともに（３）式に入力することにより、特定方向注視度C_gazeを算出できる。
また、所定の閾値は、正規利用者が撮像部を直視しているか否かの判定に用いられる閾値と同じ値、例えば4に設定される。 Furthermore, the emergency behavior detection unit may detect the emergency behavior of the authorized user based on behavior other than the behavior in which the authorized user looks directly at the imaging unit. For example, the emergency action detection unit detects that the authorized user is gazing at a specific place, for example, an entrance of a store for a certain period, based on the orientation of the authorized user's face detected from the monitoring image. In this case, it may be determined that the regular user is taking an emergency action. In this case, if the specific direction gaze degree C _gaze calculated by the following formula is equal to or greater than a predetermined threshold, the emergency action detection unit determines that the authorized user is facing the direction where the specific location is.

Here, the angle ψ is the yaw angle shown in FIG. 2, and is the rotation angle in the left-right direction (around the X axis) with respect to the face orientation in the face-to-face state (all rotation angles are 0). Further, the angle θ is a pitch angle, and is a rotation angle in the vertical direction (around the Y axis) with respect to the face direction in the directly-facing state. In addition, the angles ψ ₀ and θ ₀ are a yaw angle and a pitch angle, respectively, when the authorized user faces in a certain direction. The angles ψ ₀ and θ ₀ representing the direction in which the specific place is present have different values depending on the position where the regular user is located. On the other hand, the emergency action detection unit can specify the position of the authorized user in the monitoring area at the position where the authorized user is shown on the monitoring image. Therefore, the storage unit stores the angles ψ ₀ and θ ₀ in advance in association with each pixel on the monitoring image. The emergency action detection unit reads the angles ψ ₀ and θ ₀ corresponding to the pixel where the center of gravity of the head region where it is determined that the regular user is captured from the storage unit, and is detected by the face direction detection unit. The specific direction gaze degree C _gaze can be calculated by inputting the yaw angle ψ representing the orientation of the regular user's face and the pitch angle θ into the equation (3).
Further, the predetermined threshold is set to the same value as the threshold used for determining whether or not the authorized user is directly looking at the imaging unit, for example, 4.

  Furthermore, when the emergency call device has a plurality of image capturing units, the emergency action detecting unit is configured to, for each monitoring image captured by each image capturing unit, based on the face orientation of the authorized user detected from the monitoring image. It may be determined whether the authorized user is directly looking at the imaging unit. The emergency action detection unit may determine that the authorized user is taking an emergency action when it is determined that the authorized user is directly looking at the plurality of imaging units alternately multiple times.

  Further, the emergency action detection unit may determine whether or not the authorized user is taking an emergency action using information other than the orientation of the authorized user's face. Similarly, the suspicious behavior detection unit may determine whether or not the unauthorized user is performing suspicious behavior using information other than the orientation of the unauthorized user's face.

For example, according to another embodiment, the emergency call device detects the iris and line-of-sight direction of a regular user or a non-regular user based on a monitoring image, and matches a regular user or a non-regular user, Used to determine the presence or absence of behavior or the presence or absence of suspicious behavior.
In this case, the person region detection unit extracts an eye region and a pupil region from the monitoring image. The extraction of the eye region and the pupil region can be executed using, for example, any of the above-described face feature point extraction methods.
When determining whether the person shown in the monitoring image is a regular user or a registered non-regular user based on the iris, the storage unit stores the face image of the regular user or the registered non-regular user. Instead, the pupil image of the regular user or the non-regular user is stored. The abnormal behavior detection unit has a matching unit that performs iris matching instead of the face matching unit, and the matching unit includes a pupil region extracted from the monitoring image and a regular user stored in the storage unit or The degree of coincidence is calculated by performing pattern matching with the pupil image of the non-regular user. Then, if the maximum value of the matching degree is equal to or greater than a predetermined threshold, the matching unit determines that the person shown in the monitoring image is a regular user or a non-regular user corresponding to the maximum matching degree. On the other hand, if the maximum value of the matching degree is less than a predetermined threshold, the collation unit determines that the person shown in the monitoring image is neither a regular user nor a registered non-regular user.

In addition, the emergency action detection unit and the suspicious action detection unit, for example, when detecting a state in which the line-of-sight direction is greatly inclined in any of the up, down, left, and right directions with respect to the front of the face, Determine that you are taking suspicious behavior. For example, an eye region is approximated as an ellipse, and an elliptical region having a major axis and a minor axis that are 40% of the major axis and minor axis, respectively, and whose center of gravity coincides with the center of gravity of the eye region is defined as a normal visual line region. At this time, the emergency behavior detection unit and the suspicious behavior detection unit increment the emergency behavior counter and the suspicious behavior counter by 1, respectively, if the center of gravity of the pupil region is not included in the normal line-of-sight region.
Note that the emergency behavior detection unit and the suspicious behavior detection unit may use other gaze detection directions as disclosed in, for example, Japanese Patent Laid-Open No. 06-6786, and use the gaze direction of a regular user or a non-regular user. May be determined.

  Furthermore, the emergency call device is configured so that the regular user or the non-regular user directs his / her line of sight to the imaging unit based on both the face direction and the gaze direction of the regular user or non-regular user detected from the monitoring image. Whether or not an emergency action or a suspicious action is taken may be determined by determining whether or not there is. In this case, the face direction detection unit sets a value obtained by adding the horizontal angle of the gaze direction with respect to the front direction of the face to the yaw angle representing the face direction as the gaze direction yaw angle ψ ′, and determines the face direction. A value obtained by adding the angle in the vertical direction of the line-of-sight direction with respect to the front direction of the face to the pitch angle to be represented is defined as the line-of-sight direction pitch angle θ ′. Then, the direct visibility in the visual line direction is calculated by inputting the visual line direction yaw angle ψ ′ and the visual line direction pitch angle θ ′ into the equation (1) instead of the yaw angle ψ and the pitch angle θ. Then, the emergency behavior detection unit and the suspicious behavior detection unit determine that the regular user or the non-regular user is directing his / her line of sight to the imaging unit if the directivity in the gaze direction is equal to or greater than the threshold for the direct vision. .

Furthermore, the suspicious behavior detection unit checks the line-of-sight direction of the unauthorized user for a plurality of monitoring images, and the number of times the line-of-sight direction changes by a predetermined angle (for example, 90 °) or more is a predetermined number of times within a certain period The non-regular user may determine that he / she is taking suspicious behavior. At that time, the suspicious behavior detection unit also examines the orientation of the face, and if the orientation of the face hardly changes and only the line of sight changes as described above, the unauthorized user takes suspicious behavior. It may be determined that
Furthermore, the emergency behavior detection unit and the suspicious behavior detection unit may detect the emergency behavior and the suspicious behavior using different types of information. For example, the emergency behavior detection unit may detect the emergency behavior based on the face orientation of the authorized user, and the suspicious behavior detection unit may detect the suspicious behavior based on the direction of the line of sight.
Further, the emergency call device may not have the suspicious behavior detection unit and the suspicious behavior registration unit when only the fact that the authorized user has taken the emergency behavior is the detection target.
As described above, those skilled in the art can make various modifications in accordance with the embodiment to be implemented within the scope of the present invention.

DESCRIPTION OF SYMBOLS 1 Emergency call device 2 Imaging part 3 Interface part 4 Storage part 5 Output part 6 Abnormal action detection part 61 Person area detection part 62 Face direction detection part 63 Face collation part 64 Emergency action detection part 65 Suspicious action detection part 66 Suspicious person registration part 67 Report Department

Claims (5)

A collation unit for determining whether or not an image of a person included in a monitoring image obtained by capturing a monitoring area is an image of a regular user registered in advance;
An emergency action detection unit for determining whether the person is the authorized user and the person has taken a predetermined emergency action in the monitoring image;
A notification unit for making an emergency call when it is determined that the emergency action detection unit has taken the emergency action;
An emergency call device characterized by comprising: A suspicious behavior detection unit for determining whether or not the person has taken a predetermined suspicious behavior in the monitoring image when the collation unit determines that the person is an unauthorized user who is not the regular user; Have
The emergency reporting device according to claim 1, wherein the reporting unit issues an alarm different from the emergency reporting when it is determined that the suspicious behavior detecting unit has taken the suspicious behavior. A face direction detection unit for detecting a face direction of a person shown in the monitoring image;
The emergency action detection unit detects that the person has taken an emergency action when the face direction detection unit detects the face direction of the person in the first direction in a predetermined number of consecutive times in the monitoring images taken in time series. The emergency call device according to claim 1, wherein the emergency notification device is determined. A face direction detection unit for detecting a face direction of a person shown in the monitoring image;
The suspicious behavior detection unit determines that the person has taken suspicious behavior when the face orientation detection unit detects the face orientation of the person in the second direction over a predetermined number in the monitoring images taken in time series. The emergency call device according to claim 2. The face direction detection unit calculates a direct visibility representing a degree that the person is facing a front face in the monitoring image from the face direction;
The emergency call device according to claim 3, wherein the emergency action detecting unit sets the face direction in the first direction when the direct visibility is equal to or greater than a predetermined threshold.

Images