JP6451418B2 - Gaze target determination device, gaze target determination method, and gaze target determination program - Google Patents

Description

  The present invention relates to a technique for processing a frame image captured by an imaging device such as a video camera and setting a gaze target from the behavior of a person being captured.

  Conventionally, in places where a large number of unspecified people gather, such as stations, shopping centers, and downtowns, captured images of installed video cameras and the like (hereinafter simply referred to as cameras) are processed, and humans and the like that are captured (hereinafter referred to as cameras) , Simply referred to as the subject). Specifically, the frame image captured by the camera is processed, the behavior of the subject is analyzed, and it is determined whether or not the behavior is unique. For example, Patent Document 1 discloses a technique for detecting a behavior that moves while wobbling based on the number of times of turn of a subject included in a frame image.

JP-A-6-213632

Detecting a behavior that moves while wobbling is effective for finding a poor physical condition and preventing accidents of drunks. However, even if a behavior with a large number of direction changes is detected as in the technique disclosed in Patent Document 1 described above, for example, avoiding others in an area with high congestion such as a crowded or rush hour station. However, there is a risk that even a moving subject may be treated as a subject having a unique behavior.

  SUMMARY OF THE INVENTION An object of the present invention is to provide a technique capable of improving the accuracy of determining a target person who takes a peculiar behavior such as moving while wobbling as a gaze target.

  The gaze target determination apparatus of the present invention is configured as follows in order to solve the above-described problems and achieve the object.

The gaze target determination device processes the frame image captured by the imaging device and extracts the captured object. Then, the gaze target determination device acquires a trajectory of the object corresponding to the position and time at which the extracted object was imaged. In addition, the gaze target determination device detects the number of times that the change in the movement direction of the extracted object is unique based on the acquired trajectory of the object. Further, the speed parameter of the extracted object is detected based on the acquired trajectory of the object. Then, the gaze target determination device determines whether or not to set the extracted object as a gaze target based on the number of times of peculiarity (hereinafter referred to as the number of wobbles) and the detected speed parameter.

  In other words, the gaze target determination device can analyze the behavior of the subject by adding the speed of the subject to the number of times the subject's wobbling extracted as the object. It is known that, when a drunk person or a person with poor physical condition moves while wandering, the speed becomes extremely fast, conversely extremely slow, and the speed variation becomes large. On the other hand, it is known that in a behavior in which a standard subject moves while avoiding others, the movement speed is generally stable and easily falls within a predetermined range. In the gaze target determination device described above, a unique behavior that moves while the subject is wobbling is detected by paying attention to the speed of the subject. As a result, it can be detected more accurately.

Further, gaze target determining unit, as the evaluation score the product of the variability number the coefficients Fu, detects the coefficient based on the speed parameter, when the evaluation score is greater than the threshold value, sets an object to gaze target. Therefore , it is possible to perform an analysis that appropriately considers the speed by correcting the evaluation score with the coefficient related to the speed parameter while analyzing the behavior mainly based on the number of times of wobbling.

  The gaze target determination device may determine whether or not the speed parameter is greater than a threshold value, and may increase the coefficient when the speed parameter is greater than the threshold value. This makes it possible to more accurately detect a target person to be a gaze target, taking into account the behavior of drunk customers and those with poor physical condition, sometimes with extremely high speed.

  The gaze target determination device may determine whether or not the speed parameter is smaller than the threshold value, and may increase the coefficient when the speed parameter is smaller than the threshold value. Accordingly, it is possible to more accurately detect a target person to be a gaze target in consideration of a behavior of a drunk customer or a person with poor physical condition whose speed is sometimes extremely slow.

  The coefficient may be a function depending on the speed parameter. For example, in order to increase the weight for the behavior of extremely high speed as described above, this function is preferably increased in accordance with the speed parameter. On the other hand, in order to increase the weight for the behavior in which the speed becomes extremely slow as described above, it is preferable that this function is reduced in accordance with the speed parameter.

  Further, the coefficient may be a function corresponding to the variance of the speed parameter. In this case, as described above, it is possible to increase the weight for the behavior having a large variation in speed, and to detect the target person to be a gaze target.

  Further, the condition for determining the wobbling may include that the amount of change in the moving direction of the object per time exceeds a threshold value. Further, the condition for determining the wobbling may include that the change in the moving direction of the object is reversed. The gaze target determination device may further include a notification unit that notifies the object set as the gaze target.

  Moreover, the gaze target determination method of this invention performs the same step as the process of each part with which the above-mentioned gaze target determination apparatus is provided.

  In addition, the gaze target determination program of the present invention is installed in a computer, and causes the computer to execute the same steps as the processing of each unit included in the gaze target determination device described above.

  According to the present invention, it is possible to improve the accuracy of determining a target person, who is extracted from a frame image and moves while wobbling, as a gaze target.

It is a block diagram which shows the structure of the principal part of a gaze target determination apparatus. It is a figure which illustrates a subject's locus. It is a figure which illustrates and demonstrates the method of detecting the singular point and speed parameter of the change of a moving direction from a locus | trajectory of a subject. It is a flowchart which shows operation | movement of an image process part. It is a flowchart at the time of updating locus data. It is a flowchart shown at the time of detecting the number of times of wobbling. It is a flowchart which shows the operation | movement at the time of detecting a coefficient based on a speed parameter. It is a flowchart which shows another operation | movement at the time of detecting a coefficient based on a speed parameter.

  Hereinafter, an embodiment of a gaze target judging device which is an embodiment of the present invention will be described.

  FIG. 1 is a block diagram showing a configuration of a main part of a gaze target determining apparatus according to an embodiment of the present invention. The gaze target determination apparatus 1 according to this embodiment processes a frame image captured by a video camera 2 installed as a surveillance camera. The video camera 2 captures a frame image of about several tens of frames (for example, 30 frames) per second and outputs the frame image to the gaze target determination device 1. For example, the video camera 2 images a monitoring target area such as a ticket gate of a station or a station platform. The gaze target determination device 1 has a function of extracting a person or an object from the frame image as a target person and, if the extracted target person is a person, determining the behavior of the person.

  The gaze target determination device 1 includes a control unit 11, a display unit 15, an image processing unit 13, a storage unit 14, and a communication unit 12. The control unit 11 is a functional block composed of a CPU or the like, and controls other processes other than image processing. The storage unit 14 is a functional block including a memory, a hard disk, and the like, and stores various types of information. The communication unit 12 is a functional block including a connection connector and a communication interface, and communicates with other external devices other than the video camera.

  The image processing unit 13 receives a frame image captured by the video camera 2 and processes the frame image. The image processing unit 13 includes an extraction unit 13A, a change detection unit 13B, a speed detection unit 13C, and a gaze target determination unit 13D.

  The extraction unit 13A processes the frame image input from the video camera 2 and extracts a target person such as a person or an object being captured. Specifically, the extraction unit 13A extracts a pixel block (target person) that can specify a person or an object from the input frame image, and gives an ID to the target person. The ID is a unique value that can identify the target person. Then, the extraction unit 13A extracts coordinate data 14B such as the head of the subject. The extraction unit 13A links the extracted coordinate data 14B to the ID of the target person and causes the storage unit 14 to store it.

  The storage unit 14 stores a plurality of coordinate data 14B for each registered ID. Each of the plurality of coordinate data 14B for each ID stored in the storage unit 14 is based on a frame image captured at a constant time interval. That is, the series of coordinate data 14B stored for each ID corresponds to trajectory data 14A indicating the trajectory of the subject corresponding to the position and time at which the subject was extracted by the extraction unit 13A. Therefore, the storage unit 14 corresponds to a trajectory acquisition unit referred to in the present invention.

  The trajectory data 14A is data capable of grasping the trajectory of the subject on the subject's frame image or on the actual plane. FIG. 2A is a diagram showing an example in which the trajectory data 14A for each ID is drawn in the coordinate system of the frame image, and FIG. 2B is the trajectory data 14A for each ID in the coordinate system of the real plane. It is a figure which shows the example which coordinate-transformed and drawn with the coordinate system of the real plane. The trajectory data 14A stored in the storage unit 14 by the extraction unit 13A may be based on the coordinate system of the frame image, or may be based on the coordinate system on the real plane. However, in order to improve the detection accuracy of the change in the movement direction of the subject described later, it is desirable to use the coordinate data 14B of the coordinates on the real plane. For this reason, in the following description, unless otherwise specified, the coordinate data 14B is coordinate-transformed into a coordinate system on a real plane. Various processes based on the coordinate data 14B described below may be based on the coordinate system of the real plane or the coordinate system of the frame image.

The change detection unit 13B detects a change in the movement direction of the subject based on the trajectory data 14A for each ID. FIG. 3A is a diagram illustrating an example of a method for detecting a change in the moving direction of the subject from a plurality of coordinate data 14B included in the trajectory data 14A. On the left side of the figure , the trajectory of the subject corresponding to certain trajectory data 14A is shown.

  In the method described here, first, the time-change angle Δθ in the moving direction of the subject is detected. The moving direction of the subject is a vector that connects any two points of the coordinate data 14B. Therefore, the angle Δθ corresponds to the angle formed by the vectors of the two movement directions. In the example shown in the drawing, the moving direction of each coordinate data 14B is a vector from the previous coordinate data 14B separated by 3 frames, and the angle formed by the two moving direction vectors separated by 5 frames between each other is set as the angle of the subject. The time change angle Δθ in the moving direction is used.

  The change detection unit 13B can detect a change in the movement direction of the subject by processing the trajectory data 14A and performing a calculation corresponding to the above. The method for detecting the change in the movement direction of the subject as described above is one example of detecting the change in the movement direction of the subject from the trajectory data 14A in the coordinate system on the real plane. The change in the moving direction of the subject may be detected by any detection method.

  Further, the change detection unit 13B detects a singular point of the detected change in the moving direction of the target person, and detects the number of times the singular point is generated as the number of times of fluctuation C. FIG. 3B is a diagram illustrating an example of a method for detecting a singular point based on the angle Δθ. FIG. 3B is a graph illustrating the transition of the angle Δθ detected from the coordinate data 14B included in the movement trajectory of a certain subject. Further, in FIG. 3B, a part detected as a singular point using several conditions is surrounded by a circular display.

  In the method described here, an angle Δθ that satisfies the following three conditions is detected as a singular point.

Condition 1. The absolute value of the angle Δθ in the current frame is greater than the threshold value Ds2,
Condition 2. From the previous frame detected as a singular point to the current frame is more than the predetermined number of frames,
Condition 3. Between the timing when the singular point was detected last time and the current frame, a change in the angle Δθ in the direction opposite to the current frame (opposite polarity) occurred at a level where the absolute value exceeds the threshold value Ds1 ( However, this condition is not considered in the first singularity.)

  By using such conditions, it is possible to detect a singular point of change in the moving direction of the subject based on the angle Δθ of the time change in the moving direction of the subject.

Change detecting section 13B, by processing the trajectory data 14A performs arithmetic operations corresponding to the above, detecting the singular point from the change in the moving direction of the subject, that detect the number of times C wobble has been detected singularities be able to. The detection method of the singular point as described above (wobble number counting method) is merely illustrative, conditions and the imaging area of the frame image, the subject, in accordance with the circumstances such behavior, the appropriate singularity The detection method and detection conditions may be determined.

  Further, the speed detection unit 13C illustrated in FIG. 1 detects the speed parameter V of the subject based on the trajectory data 14A for each ID, and stores the detected speed parameter V in the storage unit 14.

  FIG. 3C is a diagram illustrating an example of a method for detecting the speed parameter V of the subject from a plurality of coordinate data 14B included in the trajectory data 14A. On the left side of the figure, the trajectory of the subject corresponding to certain trajectory data 14A is shown.

  The speed parameter V of the subject corresponds to the distance from the coordinate data 14B of the frame preceding the coordinate data 14B of the reference frame by a unit time to the coordinate data 14B of the reference frame. The example shown corresponds to the coordinate system of the real plane, and the distance between the coordinate data 14B in the n frame and the coordinate data 14B in the n-3 frame is set as the target person, assuming that the unit time of the speed parameter is 3 frames. The speed parameter. Therefore, the speed detection unit 13C illustrated in FIG. 1 can detect the speed parameter V of the subject by processing the trajectory data 14A and performing a calculation corresponding to the above.

  Note that the method for detecting the speed parameter of the subject as described above is one example of detecting the speed parameter of the subject from the locus data 14A of the coordinate system on the real plane. You may detect with such a detection method. Further, any index of the speed parameter V of the subject to be detected may be detected as long as the index correlates with the actual speed of the subject. For example, when the velocity parameter is obtained from the coordinate data 14B in the frame image coordinate system, the vertical distance and the horizontal distance are different in the frame image coordinate system. It is possible to detect the longitudinal velocity parameter and the lateral velocity parameter of the subject from the coordinate difference in the direction and the coordinate difference in the lateral direction, and use the larger one as the velocity parameter of the subject. Good.

  Further, the gaze target determination unit 13D illustrated in FIG. 1 detects the coefficient K based on the speed parameter V of the subject detected by the speed detection unit 13C. Then, the gaze target determination unit 13D detects the evaluation score S based on the coefficient K and the number of fluctuations C detected by the change detection unit 13B. Then, the gaze target determination unit 13D determines whether or not the evaluation score S is greater than the threshold value, and determines that the behavior of the subject corresponds to the wandering behavior when the evaluation score S is greater than the threshold value. The target person is set as a gaze target. When the gaze target is set, the gaze target determination unit 13 </ b> D notifies the external notification device of the gaze target person via the communication unit 12, or notifies the gaze target person from the display unit 15. Or do. Therefore, these communication part 12 and the display part 15 are corresponded to the alerting | reporting part said by this invention, respectively.

  Since the gaze target determination device 1 is configured as described above, the gaze target determination device 1 has a good accuracy in the wobbling behavior that is characterized by both the speed parameter V of the subject and the number of times C the subject wanders. It is possible to detect a target person who has taken a peculiar behavior that moves while wobbling and can be detected more accurately as a gaze target.

  Hereinafter, the gaze target determination process using the evaluation score S in the gaze target determination device 1 will be described in more detail.

  FIG. 4 is a diagram illustrating a control flow of the image processing unit 13 in the gaze target determination device 1.

  Each time the image processing unit 13 acquires a frame image of the target area captured by the video camera 2, the image processing unit 13 updates the trajectory data 14A based on the acquired frame image (S10).

  Next, the image processing unit 13 detects the number of fluctuations C for each subject based on the updated trajectory data 14A (S20).

  Further, the image processing unit 13 detects the coefficient K for each subject based on the updated trajectory data 14A (S30).

  Next, the image processing unit 13 detects the evaluation score S as the product of the number of times of fluctuation C and the coefficient K for each target person (S41), and uses the threshold value to determine whether the evaluation score S is greater than the threshold value. If the evaluation score S is larger than the threshold value, the target person is set as a gaze target. Specifically, the threshold Ss2 is used, and if the evaluation score S is larger than the relatively small threshold Ss2, the target person is set as a gaze target. These steps correspond to the gaze target determination step referred to in the present invention.

  Here, the image processing unit 13 further uses a threshold value Ss1 larger than the threshold value Ss2, and determines whether or not the evaluation score S of the subject set as the gaze target is larger than the threshold value Ss1. When the evaluation score S is larger than the threshold value Ss1, the image processing unit 13 determines that the corresponding target person needs urgent action, and outputs a signal that warns the target person's stagger behavior to an external device such as an alarm device. (S42). When the evaluation score S of the subject set as the gaze target is smaller than the threshold value Ss1, the image processing unit 13 outputs a signal to watch out for the wandering behavior of the subject to an external device such as an alarm device (S43). .

  Since the image processing unit 13 performs the gaze target determination process with the control flow as described above, the evaluation score S is weighted when the speed is unique for the target person with a small number of wobbles. It can be easily determined that the subject is gaze. In addition, for a subject who has a large number of wobbles, it is possible to change the weighting of the evaluation score S according to whether or not the speed is singular, thereby facilitating the determination as to whether or not the wandering behavior should be warned.

  FIG. 5 is a diagram showing a more detailed control flow for the update process (S10) of the trajectory data 14A.

  First, the image processing unit 13 extracts a pixel block (target person) that can specify a person or an object from the acquired frame image (S11). This step corresponds to the extraction step in the present invention. Next, the image processing unit 13 assigns an ID to the extracted target person (S12). The ID is a unique value that can identify the target person. At this time, it is detected using a known human body detection technique or the like whether or not the subject person estimated to be the same in the previous frame image has been extracted. When the same ID is detected, the same ID as the previous time is given to the target person, and when the same ID has not been extracted last time, a new ID is given. Further, the image processing unit 13 registers a new ID in the trajectory data 14A when the target person estimated to be the same in the previous frame image has not been extracted (S13). Next, the image processing unit 13 extracts the coordinate data 14B of the representative point from each pixel block (target person) to which the ID is assigned (S14). Next, the image processing unit 13 performs coordinate conversion to the coordinate system on the real plane with respect to the coordinate data 14B on the frame image (S15). Next, the image processing unit 13 reads the position of the subject's past coordinate data 14B from the trajectory data 14A, and performs leveling processing of the subject's movement trajectory together with the position of the subject's current coordinate data 14B ( S16).

  Then, the image processing unit 13 links the ID to the coordinate data 14B of the subject detected by these processes and stores it in the storage unit 14, thereby updating the trajectory data 14A (S17). By such processing, the image processing unit 13 can update the trajectory data 14A. This step corresponds to the trajectory acquisition step referred to in the present invention.

  FIG. 6 is a diagram showing a more detailed control flow for the detection process (S20) of the number of times of fluctuation C.

  Here, it is determined whether or not a new singular point is included in the updated trajectory data 14A, and the number of fluctuations C is counted. Specifically, the image processing unit 13 first detects a recent change in movement direction (change amount Δθ) in the updated trajectory of the target person by the method described above with reference to FIG. (S21). Next, the image processing unit 13 uses the method described above with reference to FIG. 3B or the like, and the recent change in the moving direction in the updated trajectory of the subject corresponds to a newly generated singular point. It is determined whether it is a thing (S22). Next, when determining that the image processing unit 13 corresponds to a singular point, the image processing unit 13 increases the count of the number of times of wander C (S23). By such processing, the image processing unit 13 can detect the number of times of wander C based on the trajectory data 14A. These steps correspond to the change detection step in the present invention.

  FIG. 7 is a diagram showing a more detailed control flow for the coefficient K detection process (S30).

  Here, the latest speed parameter V is detected from the updated trajectory data 14A, and it is determined whether or not the detected speed parameter V is a speed parameter generated in accordance with the behavior of the subject that fluctuates. The coefficient K is increased when it is determined that the speed parameter is correct.

  First, the image processing unit 13 detects the latest speed parameter V in the trajectory of the target subject using the method described above with reference to FIG. 3C (S31). This step corresponds to the speed detection step in the present invention.

  Next, the image processing unit 13 determines whether or not the detected speed parameter V is a predetermined speed with respect to the threshold value Vs in order to determine whether or not the detected speed parameter V corresponds to a speed that occurs in accordance with the subject's staggering behavior. It is determined whether or not the above condition is satisfied (S32).

  Next, the image processing unit 13 increases the coefficient K when it is determined that the detected speed parameter V satisfies a predetermined condition with respect to the threshold value Vs (S33), and when the condition is not satisfied, The coefficient K is set to 1 (S34).

  In this way, the image processing unit 13 increases the coefficient K based on whether or not the speed parameter V satisfies a predetermined condition. Therefore, the evaluation score S described above has the speed parameter V that satisfies the predetermined condition. When it does, it becomes high and it becomes easy for a subject to be judged as a gaze target. As described above, in the case of a drunk customer or a person with poor physical condition, the speed may become extremely fast or slow, so that the speed parameter V becomes larger than the threshold or the speed parameter V becomes smaller than the threshold. By increasing the coefficient K on the condition of the above, taking into account the behavior that the speed becomes extremely fast or slow, the behavior is detected so that the subject moves while wobbling, and the accuracy of setting the gaze target is increased. It is done.

  Further, when the coefficient K is increased as described above, the amount by which the coefficient K is increased may be the function f (V) of the speed parameter V. In this case, if the function f (V) is a function that increases as the speed parameter V increases, it is possible to increase the weighting of the behavior that drastically increases the speed of the drunk or poor physical condition. It is possible to easily set a target person who takes such behavior as a gaze target. Alternatively, if the function f (V) is a function that decreases with an increase in the speed parameter V, it is possible to increase the weighting of the behavior that drastically reduces the speed of the drunk or the poor physical condition. It is possible to make it easy to set a subject who takes a behav- ior as a gaze target.

  As the above condition (S32) and function (S33), an appropriate one may be adopted depending on the type of behavior to be considered for speed, the environment of the target area, and the like, or only one of them may be used. . The above-described various arithmetic expressions are merely examples, and may be replaced with another arithmetic expression that obtains an equivalent result, or may be replaced with a process other than the arithmetic that obtains an equivalent result. The contents of the above-described conditions and arithmetic expressions may be appropriately changed or adjusted according to the actual situation such as the target behavior and the environment of the target area.

  FIG. 8 is a control flow diagram illustrating a modification example of the coefficient K detection process (S30).

  In the modified example shown in FIG. 8, the variance A is detected together with the velocity parameter V (S31 ′), and the function f (A) of the detected variance A is set as an increase amount when the coefficient K is increased (S33 ′). In this case, for example, it is possible to increase the weighting of behavior that causes a large variation in the speed of a drunk customer or a person with poor physical condition, and it is possible to easily set a target person who takes such behavior as a gaze target. .

  By adopting the control flow as described above, the image processing unit 13 corrects the evaluation score S by the coefficient K related to the speed parameter V while analyzing the behavior mainly based on the number of times of fluctuation C, and appropriately It is possible to perform an analysis taking into account the speed.

DESCRIPTION OF SYMBOLS 1 ... Gaze target determination apparatus 2 ... Video camera (imaging apparatus)
DESCRIPTION OF SYMBOLS 11 ... Control part 12 ... Communication part 13 ... Image processing part 13A ... Extraction part 13B ... Change detection part 13C ... Speed detection part 13D ... Gaze target determination part 14 ... Storage part 14A ... Trajectory data 14B ... Coordinate data 15 ... Display part

Claims (10)

An extraction unit that processes the frame image captured by the imaging device and extracts the captured object;
A trajectory acquisition unit that acquires a trajectory of the object corresponding to the position and time at which the object extracted by the extraction unit was imaged;
Based on the trajectory of the object acquired by the trajectory acquisition unit, a change detection unit that detects the number of times the change in the movement direction of the object extracted by the extraction unit is unique;
Based on the trajectory of the object acquired by the trajectory acquisition unit, a speed detection unit that detects a speed parameter of the object extracted by the extraction unit;
A gaze target determination unit that determines whether to set the object extracted by the extraction unit as a gaze target based on the number of times detected by the change detection unit and the speed parameter detected by the speed detection unit; Bei to give a,
The gaze target determination unit uses the product of the number of times and the coefficient as an evaluation score, detects the coefficient based on the speed parameter, and sets the object as a gaze target when the evaluation score is greater than a threshold value. ,
Gaze target determination device. The gaze target determination unit determines whether or not the speed parameter is greater than a threshold, and increases the coefficient when the speed parameter is greater than the threshold.
The gaze target determination device according to claim 1 . The gaze target determination unit determines whether the speed parameter is smaller than a threshold, and increases the coefficient when the speed parameter is smaller than the threshold.
The gaze target determination device according to claim 1 . The coefficient is a function depending on the speed parameter.
The gaze target determination device according to any one of claims 1 to 3 . The coefficient is a function depending on the variance of the speed parameter.
Gaze target determining apparatus according to any one of claims 1 to 4. The change detection unit includes, as a condition for determining whether or not the change in the movement direction is singular, that the amount of change in the movement direction per time of the object exceeds a threshold value,
The gaze target determination device according to any one of claims 1 to 5 . The change detection unit includes that the change in the movement direction of the object is reversed as a condition for determining whether or not the change in the movement direction is singular.
The gaze target determination device according to any one of claims 1 to 6 . A notification unit that notifies the object set by the gaze target determination unit as a gaze target;
The gaze target determination device according to any one of claims 1 to 7 . An extraction step of processing the frame image captured by the imaging device and extracting the captured object;
A trajectory acquisition step for acquiring a trajectory of the object corresponding to the position and time at which the object extracted in the extraction step was imaged;
Based on the trajectory of the object acquired in the trajectory acquisition step, a change detection step of detecting the number of times the change in the moving direction of the object extracted in the extraction step is unique;
Based on the trajectory of the object acquired in the trajectory acquisition step, a speed detection step of detecting a speed parameter of the object extracted in the extraction step;
Based on the number of times detected in the change detection step and the speed parameter detected in the speed detection step, a gaze target determination step for determining whether to set the object extracted in the extraction step as a gaze target;
A method for determining a gaze target,
The gaze target determination step uses the product of the number of times and the coefficient as an evaluation score, detects the coefficient based on the speed parameter, and sets the object as a gaze target when the evaluation score is greater than a threshold value. Is a step,
Gaze target determination method. An extraction step of imaging device processes the frame image captured, to extract the imaged object,
A trajectory acquisition step for acquiring a trajectory of the object corresponding to the position and time at which the object extracted in the extraction step was imaged;
Based on the trajectory of the object acquired in the trajectory acquisition step, a change detection step of detecting the number of times the change in the moving direction of the object extracted in the extraction step is unique;
Based on the trajectory of the object acquired in the trajectory acquisition step, a speed detection step of detecting a speed parameter of the object extracted in the extraction step;
Based on the number of times detected in the change detection step and the speed parameter detected in the speed detection step, a gaze target determination step for determining whether to set the object extracted in the extraction step as a gaze target;
The to be executed by the computer, a gaze object determination program,
The gaze target determination step uses the product of the number of times and the coefficient as an evaluation score, detects the coefficient based on the speed parameter, and sets the object as a gaze target when the evaluation score is greater than a threshold value. Is a step,
Gaze target determination program.