JP7430362B2 - Abnormal behavior detection system - Google Patents

Description

The present invention relates to an abnormal behavior detection system for detecting abnormal behavior of a moving object shown in a video.

Conventionally, normal behavior in a predetermined range is determined based on captured time-series images, and when an action different from the normal action is detected in the predetermined range, the behavioral object that performed the action different from the normal action is identified. A technique for extraction is known (for example, see Patent Document 1).

Patent No. 6647489

However, for example, when the shooting range is wide, it is expected that many types of actions will be performed. In this case, it is difficult to determine to what extent the behavior should be determined as normal behavior (setting a threshold value), and as a result, abnormal behavior may not be detected or even behavior without a problem may be detected as abnormal behavior.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide an abnormal behavior detection system that can detect abnormal behavior of a moving object shown in a video with high accuracy.

The present invention includes a first acquisition unit that acquires a sample image photographed by a photographing means installed to photograph a predetermined range, and a first acquisition unit that acquires a sample image photographed by a photographing means installed to photograph a predetermined range; a first detection unit that detects in association with a first detection unit, and one or more normal behaviors at one or more positions within the predetermined range, based on a large number of the behaviors detected by the first detection unit. a second acquisition unit that acquires a first target image photographed by a photographing means installed to photograph the predetermined range; a second detection unit that detects the behavior of the first target behavior object that is displayed in association with the position of the first target behavior entity; An abnormal behavior detection system comprising: a determination unit that outputs an output indicating that when the behavior differs from the normal behavior determined at the one position; and a determination device.

According to another aspect of the present invention, there is provided an abnormal behavior detection program, an abnormal behavior detection method, a determination device, a determination program, a determination method, a learning device, a learning program, and a learning method that correspond to the abnormal behavior detection system. providing.

According to the abnormal behavior detection system of the present invention, it is possible to detect abnormal behavior of a moving object shown in a video with high accuracy.

An explanatory diagram of the usage state of the abnormal behavior detection system according to the first embodiment of the present invention Block diagram of abnormal behavior detection system according to the first embodiment of the present invention An explanatory diagram of normal behavior and abnormal behavior according to the first embodiment of the present invention Flowchart of the abnormal behavior detection system according to the first embodiment of the present invention Block diagram of an abnormal behavior detection system according to a second embodiment of the present invention Flowchart of abnormal behavior detection system according to the second embodiment of the present invention

Hereinafter, an abnormal behavior detection system 1 according to a first embodiment of the present invention will be described with reference to FIGS. 1 to 4.

As shown in FIG. 1, the abnormal behavior detection system 1 is for detecting abnormal behavior of a behavioral object Z shown in a video Y (frames constituting the video in FIG. 1) photographed by a photographing means X. be. In this embodiment, a human being is employed as the action object Z, and the action object Z is simply displayed with only a skeleton for easy understanding.

As shown in FIG. 2, the abnormal behavior detection system 1 includes a learning device 2 and a determination device 3, and the determination device 3 detects abnormal behavior by referring to the data learned by the learning device 2. do.

The learning device 2 includes a first storage section 21, a first acquisition section 22, a first detection section 23, and a determination section 24.

The first storage unit 21 stores "joint identification criteria," "behavior identification criteria," and "behavior identification criteria."

The "joint identification standard" is for identifying multiple joints A (in FIG. 1, the neck, right elbow, left elbow, waist, right knee, and left knee) of the behavioral object Z, and for each joint A, It shows the shape, direction, size, etc. for identifying each.

The “behavior identification criteria” includes the “basic posture” of various variations (“walking”, “standing up”, etc.) of the behaving entity Z, the “range of motion of each joint A”, and the “range of motion of each joint A” of each behaving entity Z. This shows the distance between the two.

The "behavior identification standard" indicates the movement of each joint when the behavioral object Z performs a predetermined action, and in this embodiment, the "action identification standard" for various types of actions is stored. has been done.

The first acquisition unit 22 acquires a sample image Y1 photographed by a photographing means X installed to photograph a predetermined range. In this embodiment, an example will be described in which the entire photographing range by the photographing means X is set as the predetermined range.

The first detection unit 23 detects the behavior of the sample action object Z1 shown in the sample video Y1 in association with the position of the sample action object Z1.

In detecting the sample behavioral object Z1, first, the sample behavioral object Z1 shown in the sample video Y1 is specified.

In this embodiment, after detecting a plurality of joints A that correspond to the "joint identification criteria" stored in the first storage unit 21, one sample action object is detected by referring to the "behavior identification criteria". A plurality of joints A included in Z1 are specified. In the example of FIG. 1, joints A1-A6 are identified as joints A included in one sample behavior object Z1, and one behavior entity Z is identified as existing.

After identifying the sample behavioral object Z1, the behavior of the sample behavioral object Z1 is detected. In this embodiment, when a movement of each joint that corresponds to the "behavior identification criterion" is detected, it is detected that "the corresponding action was performed".

As the position of the sample action object Z1, it is possible to use a position in real space.

Specifically, it is possible to estimate the position in real space by taking into consideration the plane coordinates on the screen, the coordinates in the depth direction, the size of the sample behavioral object Z1, and the like. As the coordinates in the depth direction, it is possible to use known depth estimation or the like.

Note that the "position" in this embodiment is not limited to one coordinate, but may also be a certain range. For example, if the screen is divided into multiple parts and a first target action object Z2 exists in one divided screen, the behavior of the first target action object Z2 is detected in association with the position of the first divided screen. It is possible to do so.

Further, in the present embodiment, the first detection unit 23 detects the behavior of the sample behavioral object Z1 in association with the body orientation of the sample behavioral object Z1.

As a method for specifying the body orientation of the sample behavioral object Z1, for example, it may be specified by considering the positions and directions of a plurality of joints A, or by detecting the face or hair.

The determining unit 24 determines one or more “normal behaviors” at one or more positions within a predetermined range based on the large number of behaviors detected by the first detecting unit 23 .

For example, as shown in Figure 1, when a vending machine is shown in the sample video Y1, in the position in front of the vending machine, there are options such as "insert a coin (into the coin slot)" and "(product selection button)". ), ``Press (the item) from the item take-out slot,'' ``Operate (the return lever),'' and ``Take out (the coin from the coin return slot)'' actions are determined as ``normal actions.'' It is possible that Furthermore, a series of a plurality of actions performed within a predetermined period of time as described above may be determined as a "normal action."

Furthermore, for example, the action of "taking out a product from the product outlet" may be accompanied by the action of "squatting," so in front of a vending machine, the action of "squatting" is considered a "normal action." ” (Figure 3(a)). On the other hand, in other places, the action of "squatting" may not be determined as a "normal action" (FIG. 3(b)).

Further, in the present embodiment, the determining unit 24 determines the "normal behavior" by also considering the direction of the moving object.

For example, in front of a vending machine, the behavior of ``squatting down while facing the vending machine'' is determined as a ``normal behavior'' (Figure 3(a)), and the behavior of ``squatting down facing the vending machine'' is determined as a ``normal behavior'' (Figure 3(a)), and It is conceivable that the action of "squatting" is not determined to be a "normal action" (FIG. 3(c)).

Note that "normal behavior" can be determined based on various criteria, but for example, a behavior that has a proportion equal to or higher than a predetermined (threshold) among all detected behaviors may be determined as "normal behavior." It is possible to do so.

The “normal behavior” determined by the determining unit 24 is stored in the first storage unit 21.

The determination device 3 includes a second storage section 31 , a second acquisition section 32 , a second detection section 33 , and a determination section 34 .

Similar to the first storage unit 21, the second storage unit 31 stores “joint identification criteria”, “behavior identification criteria”, and “behavior identification criteria”.

The second acquisition unit 32 acquires the first target image Y2 photographed by the photographing means X installed to photograph the predetermined range. In this embodiment, it is assumed that the first target image Y2 is photographed by the photographing means X at substantially the same angle of view and magnification as that of the first acquisition section 22.

The second detection unit 33 detects the behavior of the first target action object Z2 shown in the first target image Y2 in association with the position of the first target action object Z2. In this embodiment, it is assumed that detection is performed using a method similar to that of the first detection unit 23.

When the behavior of the first target behavioral object Z2 detected at one position is different from the "normal behavior" determined at the one position, the determination unit 34 performs an "output indicating that." That is, when an inappropriate "abnormal behavior" is detected at the one location, an alert or the like will be issued.

Here, it is possible that the detected behavior is appropriate at one location but not at another location, and in such a case, the behavior is uniform at all locations within a predetermined range. If abnormal behavior is determined based on "normal behavior," abnormal behavior may not be detected or even behavior without a problem may be detected as abnormal behavior.

However, in this embodiment, as described above, one or more "normal actions" at one or more positions within a predetermined range are determined by learning, and the first When the behavior of the target behavior object Z2 is different from the "normal behavior" determined at the one position, "an output indicating this" is performed.

Therefore, for example, in front of a vending machine, in response to the action of "squatting down" that accompanies the action of "taking out a product (from the product outlet)", no "output indicating that action" will be made (see Figure 3). a)), "output indicating that" will be performed at other locations (FIG. 3(b)).

With such a configuration, it is possible to detect abnormal behavior of the first target behavioral object Z2 shown in the first target video Y2 with high accuracy. Furthermore, in detecting abnormal behavior, it is only necessary to detect the first target behavior object Z2, and there is no need to detect other objects such as a vending machine or to associate the first target behavior object Z2 with other objects. This prevents the load on the computer from increasing.

Note that the "output indicating this" may be simply outputting an alert, or may be output by specifying the first target behavioral object Z2 that has performed an action different from the "normal action". Further, it is preferable that the "output indicating this" be automatically transmitted to the registered information terminal (security company information terminal, personal mobile information terminal, etc.).

Furthermore, in this embodiment, the "normal behavior" is determined in consideration of the body direction of the sample behavioral object Z1, and the behavior and body of the first target behavioral object Z2 detected at one position are determined. If the direction is different from the "normal behavior" determined at the one position, "output indicating that" is performed.

For example, in front of a vending machine, if you face the vending machine and ``squat'', no ``output indicating that effect'' will be performed (Figure 3 (a)), and the vending machine is In response to the action of "squatting" while facing the opposite side, "an output indicating that effect" will be performed (FIG. 3(c)). This is because the behavior of ``squatting'' while facing away from the vending machine is considered to be an abnormal behavior such as retention.

With such a configuration, it becomes possible to detect abnormal behavior of the first target behavioral object Z2 shown in the first target video Y2 with higher accuracy.

Furthermore, in the present embodiment, a series of a plurality of actions of the sample action object Z1 detected by the first detection unit 23 is determined as "normal action", and the first action detected at one position is determined as "normal action". Even when the series of multiple actions of the target behavioral object Z2 is different from the "normal behavior (normal series of multiple actions)" determined at the one position, "output indicating that fact" is performed.

For example, in front of a vending machine, "insert a coin (into the coin slot)" → "press (the product selection button)" → "take out (the product from the product takeout slot)" → "operate (the return lever)" If the user does not perform one of the steps ``take out the coin from the coin return slot'' or repeatedly performs either of the steps, an ``output indicating this'' will be performed.

According to such a configuration, it becomes possible to detect abnormal behavior of the first target action object Z2 shown in the first target video Y2 with higher accuracy.

Next, detection of abnormal behavior by the abnormal behavior detection system 1 will be explained using the flowchart of FIG.

Detection of abnormal behavior is performed in two stages: a "learning stage" and a "detection stage".

(1) Learning stage

In the learning stage, first, when a sample video Y1 photographed in a predetermined range is acquired (S1), the behavior of the sample behavioral object Z1 shown in the sample video Y1 is detected in association with the position of the sample behavioral object Z1. (S2). At this time, it is preferable to further detect the orientation of the body of the sample behavioral object Z1 and link it to the behavior of the sample behavioral object Z1.

Note that it is preferable that these S1 and S2 be performed on a plurality of different sample videos Y1 so that a large number of actions are detected.

Next, one or more "normal actions" at one or more positions within a predetermined range are determined based on the large number of actions detected in S2 (S3). At this time, it is preferable that the normal behavior is determined in consideration of the body orientation of the sample behavioral object Z1.

(2) Detection stage

Subsequently, in the detection stage, first, when the first target video Y2 photographed in a predetermined range is acquired (S4), the behavior of the first target behavioral object Z2 shown in the first target video Y2 is It is detected in association with the position of the first target action object Z2 (S5). At this time, it is preferable to further detect the body orientation of the first target behavioral object Z2 and associate it with the behavior of the first target behavioral object Z2.

Subsequently, if the behavior of the first target behavioral object Z2 detected at one position in S5 is different from the "normal behavior" determined at the one position (S6: YES), "indicate that "Output" is performed (S7). At this time, it is preferable that the orientation of the body of the sample behavioral object Z1 is also taken into account to determine whether or not the behavior is different from "normal behavior."

As explained above, in the abnormal behavior detection system 1 according to the present embodiment, one or more "normal behaviors" at one or more positions within a predetermined range are determined by learning, and If the detected behavior of the first target behavioral object Z2 is different from the "normal behavior" determined at the one position, "an output indicating that" is performed.

According to such a configuration, it becomes possible to detect the abnormal behavior of the first target action object Z2 shown in the first target video Y2 with high accuracy. Furthermore, in detecting abnormal behavior, it is only necessary to detect the first target behavior object Z2, and there is no need to detect other objects such as a vending machine or to associate the first target behavior object Z2 with other objects. This prevents the load on the computer from increasing.

Furthermore, in the abnormal behavior detection system 1 according to the present embodiment, the "normal behavior" is determined by taking into account the body orientation of the sample behavioral object Z1, and the first target behavioral object detected at one position is If Z2's behavior and body orientation are different from the "normal behavior" determined at the one position, "output indicating this" is performed.

According to such a configuration, it becomes possible to detect abnormal behavior of the first target action object Z2 shown in the first target video Y2 with higher accuracy.

Further, in the abnormal behavior detection system 1 according to the present embodiment, a series of a plurality of behaviors of the sample behavior object Z1 detected by the first detection unit 23 is determined as "normal behavior", and If the series of multiple behaviors of the detected first target behavior object Z2 is different from the "normal behavior (normal series of multiple behaviors)" determined at the one position, an output indicating that. "I do.

According to such a configuration, it becomes possible to detect abnormal behavior of the first target action object Z2 shown in the first target video Y2 with higher accuracy.

Next, an abnormal behavior detection system 10 according to a second embodiment of the present invention will be described with reference to FIGS. 5 and 6.

As described above, in the first embodiment, the determination device 3 can detect “abnormal behavior” by referring to the data learned by the learning device 2, but it is also possible to detect “normal behavior” or The "normal trajectory" may change as the shooting environment changes (for example, the layout changes due to an event, the shooting range of the camera shifts, etc.). In that case, even if the determined "normal behavior" or "normal trajectory" is used to detect "abnormal behavior", the detected "abnormal behavior" may not be suitable for the shooting environment. do not have.

Therefore, in the present embodiment, the control device 4 determines whether or not a change has occurred in the photographing environment, and when a change has occurred, the control device 4 causes the determination unit 24 to resume the "normal action" or "normal trajectory". Make a decision (relearn).

In detail, the control device 4 includes a third acquisition section 41, a third detection section 42, and a control section 43, as shown in FIG.

The third acquisition unit 41 acquires the second target image Y3 photographed by the photographing means X. In this embodiment, it is assumed that a large number of second target images Y3 are acquired in association with their respective shooting times. Note that the first target video Y2 acquired by the second acquisition unit 32 may be used as the second target video Y3.

The third detection unit 42 detects a predetermined action or trajectory of the second target action object Z3 shown in the second target image Y3 in association with the position of the second target action object Z3. In this embodiment, the term is not limited to "behavior", but may also be "trajectory in addition to action" or "trajectory in place of action".

Therefore, the first detection unit 23 of the learning device 2 detects the behavior or trajectory of the sample action object Z1 in association with the position of the sample action object Z1, and the determination unit 24 detects the action or trajectory of the sample action object Z1 in association with the position of the sample action object Z1. Based on a large number of actions or trajectories, one or more "normal actions or trajectories" in a predetermined range are determined.

Similarly, the second detection unit 33 of the determination device 3 detects the action or trajectory of the first target action object Z2 in association with the position of the first target action object Z2, and the determination unit 34 detects the action or trajectory of the first target action object Z2. When the action or trajectory of the first target action object Z2 detected at a position is different from the "normal action or trajectory" determined at the one position, an output indicating that fact is performed.

The control unit 43 controls the determining unit 24 to re-determine the “normal behavior or trajectory” when the detection position of the predetermined behavior or trajectory detected by the third detection unit 42 changes by a predetermined amount or more. .

For example, as a result of continuous monitoring of the detected position of a predetermined action or trajectory, if the detected position changes by a predetermined amount or more after a certain time, it is possible to detect a change in the detection position of a predetermined action or trajectory due to changes in the layout of the shooting range, etc. Possible causes include ``the position of the camera has changed'' or ``the camera's shooting range has shifted due to an impact, etc.''. Therefore, in such a case, the control unit 43 controls the determining unit 24 to re-determine the "normal action or trajectory." "If the detected position changes by a predetermined number or more after a certain time" is a position where the predetermined action or trajectory was detected less than the first number of times within the predetermined period, and A case may be considered in which the number of detections is greater than the second number of times.

This makes it possible to appropriately detect "abnormal behavior" using "normal behavior" and "normal trajectory" that are re-determined as the photographing environment changes.

Next, relearning (re-determination) of "normal behavior or trajectory" by the abnormal behavior detection system 10 will be explained using the flowchart of FIG.

Relearning is performed after "normal behavior" or "normal trajectory" is determined in S3 of FIG. 4. Although relearning may be performed in parallel with the detection stage or separately from the detection stage, in this embodiment, an example will be described in which relearning is performed in parallel with the detection stage.

First, when the second target video Y3 is acquired (S11), the predetermined action or trajectory of the second target behavioral object Z3 shown in the second target video Y3 is determined to be the position of the second target behavioral object Z3. They are detected in association (S12).

Finally, if the detection position of the predetermined action or trajectory detected by the third detection unit 42 changes by a predetermined amount or more (S13: YES), the determination of "normal action or trajectory" is performed again (S14). .

As described above, in the abnormal behavior detection system 10 according to the present embodiment, when the detected position of a predetermined behavior or trajectory changes by more than a predetermined value, a "normal behavior or trajectory" is determined anew.

According to such a configuration, it is possible to appropriately detect "abnormal behavior" using "normal behavior" and "normal trajectory" that are newly determined as the photographing environment changes.

Note that the abnormal behavior detection system of the present invention is not limited to the embodiments described above, and various modifications and improvements can be made within the scope of the claims.

For example, in the embodiment described above, abnormal behavior was detected in front of a vending machine, but abnormal behavior may also be detected in other locations (e.g., the entrance/exit of a facility, the entrance/exit of a restroom, in front of an elevator, etc.) Of course.

Furthermore, in the above embodiment, when one of a series of actions is not performed or one of them is repeatedly performed, an "output indicating that fact" is performed. Of course, it also includes cases that do not include actions (for example, "standing all the time," "wandering around all the time," etc.).

Furthermore, in the above embodiment, "the first storage section 21 and the second storage section 31", "the first acquisition section 22, the second acquisition section 32, and the third acquisition section 41", and " The first detection unit 23, the second detection unit 33, and the third detection unit 42 were provided in the learning device 2, the determination device 3, and the control device 4, respectively, but they are not commonly used. Good too. Furthermore, the learning device 2 and the determining device 3 may be integrated. These cases are also included within the scope of the present invention.

Further, in the first embodiment, after a certain amount of data (a large number of actions) has been accumulated in the learning device 2, it is also possible to operate only the determination device 3.

Further, although it is effective to detect abnormal behavior according to the present embodiment in real time with respect to the first target video Y2, depending on the purpose of use of the abnormal behavior detection system 1, the recorded first target video It may be performed later for Y2.

Further, in the above embodiment, a human being is used as an example of the action object Z, but it is also possible to use an animal or a robot. For example, in the case of animals, it is possible to quickly deal with cases where they start defecating in a location other than the toilet.

Further, in the above embodiment, the behavior of the behavioral object is detected based on the movements of a plurality of joints, but the behavior of the behavioral object may be detected by other methods without detecting the joints.

Further, the present invention includes a first acquisition unit 22, a first detection unit 23, a determination unit 24, a second acquisition unit 32, a second detection unit 33, a determination unit 34, and a third acquisition unit as a controller. 41, the third detection unit 42, and the control unit 43, the present invention can also be applied to a program and method corresponding to the processing performed by the third detection unit 42 and the control unit 43, and a recording medium that stores the program. In the case of a recording medium, the program will be installed on a computer or the like. Here, the recording medium storing the program may be a non-transitory recording medium. A CD-ROM or the like may be considered as a non-transitory recording medium, but is not limited thereto.

1, 10 Abnormal behavior detection system 2 Learning device 3 Determination device 4 Control device 21 First storage section 22 First acquisition section 23 First detection section 24 Determination section 31 Second storage section 32 Second acquisition section 33 Second detection unit 34 Determination unit 41 Third acquisition unit 42 Third detection unit 43 Control unit X Photographing means Y Video Y1 Sample video Y2 First target video Y3 Second target video Z Behavior object Z1 Sample behavior object Z2 First target action object Z3 Second target action object

Claims (6)

a first acquisition unit that acquires a sample image photographed by a photographing means installed to photograph a predetermined range;
a first detection unit that detects the behavior of the sample behavioral object shown in the sample video in association with the position of the sample behavioral object;
a determining unit that determines one or more normal actions at one or more positions within the predetermined range based on the large number of the actions detected by the first detection unit;
a learning device having;
a second acquisition unit that acquires a first target image photographed by a photographing means installed to photograph the predetermined range;
a second detection unit that detects the behavior of the first target action object shown in the first target video in association with the position of the first target action object;
If the behavior of the first target behavioral object detected at one position is different from the normal behavior determined at the one position, a determination unit that outputs an output indicating that;
a determination device having;
Equipped with
The determining unit determines a series of a plurality of behaviors of the sample behavioral object detected by the first detecting unit as the normal behavior,
The determination unit outputs an output indicating that when a series of the plurality of actions of the first target action object detected by the second detection unit is different from the normal action determined at the one position. An abnormal behavior detection system characterized by performing the following. A program that is executed by a computer,
obtaining a sample image taken by a photographing means installed to photograph a predetermined range;
detecting the behavior of the sample behavioral object shown in the sample video by associating it with the position of the sample behavioral object;
determining one or more normal behaviors at one or more locations within the predetermined range based on the detected number of behaviors;
acquiring a first target image photographed by a photographing means installed to photograph the predetermined range;
detecting the behavior of the first target action object shown in the first target video in association with the position of the first target action object;
When the behavior of the first target behavioral object detected at one position is different from the normal behavior determined at the one position, outputting an output indicating that;
Equipped with
In the determining step, a series of a plurality of behaviors of the detected sample behavioral object is determined as the normal behavior;
In the step of outputting, when a series of a plurality of behaviors of the detected first target behavioral object is different from a normal behavior determined at the one position, an output indicating that is performed. Characteristic abnormal behavior detection program. A computer-implemented method, the method comprising:
obtaining a sample image taken by a photographing means installed to photograph a predetermined range;
detecting the behavior of the sample behavioral object shown in the sample video by associating it with the position of the sample behavioral object;
determining one or more normal behaviors at one or more locations within the predetermined range based on the detected number of behaviors;
acquiring a first target image photographed by a photographing means installed to photograph the predetermined range;
detecting the behavior of the first target action object shown in the first target video in association with the position of the first target action object;
when the behavior of the first target behavioral object detected at one position is different from the normal behavior determined at the one position, outputting an output indicating that;
In the determining step, a series of a plurality of behaviors of the detected sample behavioral object is determined as the normal behavior;
In the step of outputting, when a series of a plurality of behaviors of the detected first target behavioral object is different from a normal behavior determined at the one position, an output indicating that is performed. Characteristic abnormal behavior detection method. a first acquisition unit that acquires a sample image photographed by a photographing means installed to photograph a predetermined range;
a first detection unit that detects the behavior of the sample behavioral object shown in the sample video in association with the position of the sample behavioral object;
a determining unit that determines one or more normal actions at one or more positions within the predetermined range based on the large number of the actions detected by the first detection unit;
Equipped with
a second acquisition unit that acquires a first target image photographed by a photographing means installed to photograph the predetermined range; a second detection unit that detects the first target behavioral object in association with the position; and a normal behavior in which the behavior of the first target behavioral object detected at one position is determined at the one position. A learning device capable of communicating with a determining device having a determining unit that outputs an output indicating that when the learning device differs from the above,
The determining unit determines a series of a plurality of behaviors of the sample behavioral object detected by the first detecting unit as the normal behavior,
The determination unit outputs an output indicating that when a series of the plurality of actions of the first target action object detected by the second detection unit is different from the normal action determined at the one position. A learning device characterized by performing the following. A program that is executed by a computer,
obtaining a sample image taken by a photographing means installed to photograph a predetermined range;
detecting the behavior of the sample behavioral object shown in the sample video by associating it with the position of the sample behavioral object;
determining one or more normal behaviors at one or more locations within the predetermined range based on the detected number of behaviors;
Equipped with
The computer includes a second acquisition unit that acquires a first target image photographed by a photographing means installed to photograph the predetermined range, and a first target action that is captured in the first target image. a second detection unit that detects a body action in association with a position of the first target action object; and a second detection unit that detects a body action in association with a position of the first target action object; and a determination unit that outputs an output indicating that when the behavior differs from the normal behavior,
In the determining step, a series of a plurality of behaviors of the detected sample behavioral object is determined as the normal behavior;
The determination unit outputs an output indicating that when a series of the plurality of actions of the first target action object detected by the second detection unit is different from the normal action determined at the one position. A learning program that features the following. A computer-implemented method, the method comprising:
obtaining a sample image taken by a photographing means installed to photograph a predetermined range;
detecting the behavior of the sample behavioral object shown in the sample video by associating it with the position of the sample behavioral object;
determining one or more normal behaviors at one or more locations within the predetermined range based on the detected number of behaviors;
Equipped with
The computer includes a second acquisition unit that acquires a first target image photographed by a photographing means installed to photograph the predetermined range, and a first target action that is captured in the first target image. a second detection unit that detects a body action in association with a position of the first target action object; and a second detection unit that detects a body action in association with a position of the first target action object; and a determination unit that outputs an output indicating that when the behavior differs from the normal behavior,
In the determining step, a series of a plurality of behaviors of the detected sample behavioral object is determined as the normal behavior;
The determination unit outputs an output indicating that when a series of the plurality of actions of the first target action object detected by the second detection unit is different from the normal action determined at the one position. A learning method characterized by the following.

Images