JP2015179439A - Imaging processing device, tracing method and tracing program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve mapping accuracy of mapping photographed objects between frame images, and to enable implementation of an appropriate tracing of the photographed object.SOLUTION: An image processing device 1 has: a video image input unit 2 into which a video image having an area of object photographed by an imaging device 10 is input; and an image processing unit 3 that processes a frame image extracted at a predetermined interval from the video image input into the video image input unit 2 as a frame image of a processing object, detects the objects photographed in the frame image of the processing object, and detects a position of the detected object. Further, the image processing unit 3 is configured to map the detected objects between the frame images of the processing object, and the mapping is configured to use continuous three frame images of the processing object timewise to perform mapping of these objects photographed between three frame images of the processing object.

Description

  The present invention relates to a technique for tracking the movement of an object captured in a moving image captured by a security camera or the like.

  Conventionally, security cameras are installed in various places where an unspecified number of people gather, such as airports, stations, shopping centers, street corners, and the like.

  In addition, it is also considered to detect a person such as a suspicious person (hereinafter simply referred to as a suspicious person) who has taken a unique action by processing a moving image captured by a camera. For example, by processing a moving image captured by a camera, detecting the movement speed, movement direction, head height, etc. of the person as a feature amount of the person being imaged, whether the person is a suspicious person There is one that determines whether or not (see Patent Document 1).

  In Patent Document 1, a camera image (moving image) to be processed is not a captured image of a security camera but an in-vehicle camera.

JP2013-88870A

  However, in order to process a moving image captured by a security camera and obtain the moving speed and moving direction of the imaged person (hereinafter referred to as a target person), the person must be tracked. In order to track the person imaged in the frame image, it is necessary to accurately identify the same person imaged between the frame images.

  In particular, when a detection omission that cannot be detected by a person captured in the frame image (the person being tracked) occurs for some reason, tracking is interrupted in the frame image. Also, when multiple people overlap on the frame image, tracking cannot be performed because the identification between the frame images cannot be performed, or another person can be tracked by incorrectly associating the overlapping people on the frame image. There is.

  An object of the present invention is to provide a technique capable of improving the accuracy of identification for associating an imaged object between frame images and appropriately tracking the imaged object.

  In order to achieve the above object, the moving image processing apparatus of the present invention is configured as follows.

  A moving image obtained by capturing an image of the target area by the imaging device is input to the moving image input unit. The object detection unit processes a frame image extracted at a predetermined interval from the moving image input to the moving image input unit as a processing target frame image, and detects an object captured in the processing target frame image. The position of the detected object is detected. The object detection unit may detect a captured object using all frame images related to the moving image input to the moving image input unit as processing target frame images, or may be input to the moving image input unit. Detection of a captured object may be performed using a frame image extracted at a predetermined interval from a moving image as a processing target frame image. What is necessary is just to determine the space | interval which extracts a process target frame image according to the frame rate of the moving image input into the moving image input part. Further, this interval may be determined as a time interval or a frame interval.

  The identification unit associates the object detected by the object detection unit between the processing target frame images. The identification unit uses three processing target frame images that are temporally continuous, and associates an object captured between the three processing target frame images.

  In this configuration, for example, among the three processing target frame images that are temporally continuous, based on the similarity of the imaged object between two processing target frame images that are not positioned temporally in the middle. Objects captured between the three processing target frame images can be associated with each other. For this reason, the object detection unit detects, for some reason, an object (tracked object) captured in the processing target frame image that is temporally intermediate among the three temporal processing target frame images. ) Is detected between these three processing target frame images based on the similarity of the imaged object between the two processing target frame images that are not positioned in the middle in time. Can be identified with high accuracy.

  In addition, among the three processing target frame images that are temporally continuous, two processing targets that are not positioned in the middle in time even if the captured object overlaps the processing target frame image that is positioned in the middle in time Based on the similarity of the imaged object between the frame images, identification that associates the imaged object between the three processing target frame images can be performed with high accuracy.

  Therefore, the accuracy of identification for associating the captured object between the processing target frame images can be improved, and the captured object can be properly tracked.

  According to the present invention, the accuracy of identification for associating the imaged object between the processing target frame images is improved, and the imaged object can be properly tracked.

It is a block diagram which shows the structure of the principal part of an image processing apparatus. It is a figure which shows the function structure of an image process part. It is a figure which shows an object map. It is a figure explaining the magnitude | size and position of an object on a process target frame image. It is a figure which shows the criteria table. It is a flowchart which shows a tracking process. It is a flowchart which shows an identification process. It is a figure explaining the process which matches the person currently imaged to the process target frame image continuous in time. It is a figure explaining the process which matches the person currently imaged to three process target frame images continuous in time. It is a figure explaining the process which matches the person currently imaged to three process target frame images continuous in time. It is a figure explaining the process which matches the person currently imaged to three process target frame images continuous in time. It is a figure explaining the process which matches the person currently imaged to three process target frame images continuous in time. It is a flowchart which shows a criterion setting process. It is a figure which shows the frequency distribution which totaled the variation | change_quantity of the horizontal direction on a process target frame image. It is a flowchart which shows a speed abnormal person detection process.

  Hereinafter, an image processing apparatus according to an embodiment of the present invention will be described.

  FIG. 1 is a block diagram showing the configuration of the main part of the image processing apparatus according to this example. The image processing apparatus 1 according to this example includes a moving image input unit 2, an image processing unit 3, and an output unit 4. The image processing apparatus 1 according to this example processes a moving image captured by the imaging device 10 such as a security camera, and determines whether the moving speed of the person captured in the moving image is normal or abnormal. At the same time, the determination result is output to the host device.

  An imaging device 10 is connected to the moving image input unit 2. The imaging device 10 is a video camera that captures moving images. The imaging device 10 is a security camera installed at a place (an airport, a station, a shopping center, a street corner, etc.) where, for example, an unspecified number of people gather. A frame image related to a moving image captured by the imaging device 10 is input to the moving image input unit 2 in time series. The frame rate of the imaging device 10 may be a relatively low one of several frames / sec (3 to 5 frames / sec). The imaging area of the imaging device 10 corresponds to the target area referred to in the present invention.

  The moving image input unit 2 may be configured to input a moving image file recorded on a medium such as a USB memory, an SD memory card, or a DVD. In this case, the moving image captured by the imaging device 10 is temporarily recorded on a medium, and the moving image file recorded on the medium is input to the moving image input unit 2.

  The image processing unit 3 processes the frame image applied to the moving image input to the moving image input unit 2, tracks the movement of the person (object) being imaged, and the movement speed of the tracked person is normal. It is determined whether there is an abnormality. Further, the image processing unit 3 sets a determination criterion used for determining whether the moving speed is normal or abnormal. Details of the image processing unit 3 will be described later.

  Note that the frame images processed by the image processing unit 3 (hereinafter referred to as processing target frame images) may not be all frame images related to the moving image input to the moving image input unit 2. Specifically, the image processing unit 3 may process all frame images related to the moving image input to the moving image input unit 2 as processing target frame images, or input to the moving image input unit 2. Frame images extracted from moving images at predetermined intervals may be processed as processing target frame images (frame images not extracted here are not subjected to processing described later). The interval for extracting the processing target frame image may be determined according to the frame rate of the moving image input to the moving image input unit 2, and may be set at a time interval or a frame interval. May be.

  The image processing unit 3 includes a computer that executes the tracking method according to the present invention. The tracking program according to the present invention is installed in a computer included in the image processing unit 3.

  The output unit 4 notifies the host device of the person whose moving speed is determined to be abnormal by the image processing unit 3. Specifically, when it is determined that the movement speed of the person tracked by the image processing unit 3 is abnormal, the output unit 4 notifies the higher-level device to that effect. The host device notifies a guard or the like that a person having an abnormal movement speed has been detected, for example, by an indicator lamp or speaker installed in a security room or the like.

  FIG. 2 is a block diagram illustrating a functional configuration of the image processing unit. The image processing unit 3 includes an object detection function unit 31, an object identification function unit 32, an object map creation function unit 33, a moving speed determination function unit 34, a tabulation function unit 35, a determination criterion setting function unit 36, It has.

  The object detection function unit 31 detects a person captured in the processing target frame image as an object for each processing target frame image extracted from the frame image of the moving image captured by the imaging device 10. As described above, the processing target frame image may be all frame images related to the moving image input to the moving image input unit 2 or may be determined in advance from the moving image input to the moving image input unit 2. Frame images extracted at intervals may be used. The object detection function unit 31 detects a person being imaged by, for example, a known background difference method or frame image difference method. Since detection of objects by the background difference method or the inter-frame image difference method is known, the description thereof is omitted here. The object detection function unit 31 assigns a temporary ID to the detected person.

  The object identification function part 32 performs the identification process which matches the person imaged to these process target frame images using several process target frame images continuous in time. Details of this identification process will be described later. Based on this identification, the object identification function unit 32 determines whether the person to whom the object detection function unit 31 has given a temporary ID has already been given the real ID, or is the person who has not given the real ID at this time. Judging. The object identification function unit 32 replaces the temporary ID with the real ID if the person has already been given the real ID. On the other hand, if the person identification function unit 32 is a person who has not been given a real ID, the object identification function unit 32 replaces the temporary ID with a new real ID. In this case, the temporary ID may be substantially replaced with the real ID, or the temporary ID may be the real ID.

The object map creation function unit 33 generates the object map shown in FIG. As shown in FIG. 3, the object map includes, for each processing target frame image, the person ID detected by the object detection function unit 31, the size S on the processing target frame image, and the processing target frame image. The position (x, y) is registered in association with each other. For example, in the example of the object map shown in FIG. This shows that the person whose real ID is A123456 and the person whose A123457 are captured in the processing target frame image at t _n−1 . In addition, the person whose main ID is A123456 is that the size S on the processing target frame image is Sa1 and the position (x, y) on the processing target frame image is (x _a1 , y _a1 ). Show. Moreover, people present ID is A123457, the size S on the processing target frame image is Sb1, position on the processing target frame image (x, y) is that it was a (x _b1, y _b1) Show.

It should be noted that a frame image (a frame image that has not been extracted as a processing target frame image) related to a moving image captured by the imaging apparatus 10 is between temporally continuous processing target frame images. t _n−1 and the processing target frame image No. and t _n, it may exist between.

In addition, other processing target frame images No. t _n , the processing target frame image No. t _{n + 1} , the processing target frame image No. Also for t _{n + 2} , the person ID detected by the object detection function unit 31 is registered in association with the size S on the processing target frame image and the position on the processing target frame image. .

  FIG. 4 is a diagram illustrating a processing target frame image in which the person A and the person B are captured. As shown in FIG. 4, in this example, the height of the rectangular area surrounding the person detected by the object detection function unit 31 (vertical lengths SA and SB on the processing target frame image) is detected. Is registered in the object map as size S. Also, as shown in FIG. 4, the lower left corner of this rectangular area (the rectangular area surrounding the detected person) is registered in the object map with the detected person's position (x, y). The detected position of the person is indicated by horizontal positions Xa and Xb on the processing target frame image and vertical positions Ya and Yb on the processing target frame image.

  The area of the rectangular area surrounding the detected person may be the size of the detected person, or the lower right corner or the center of gravity of the rectangular area on the processing target frame image is detected. It may be a position.

  Therefore, the person to whom the real ID is assigned can obtain the size S on the processing target frame image and the change of the position (x, y) with the lapse of time by this object map. That is, for a person to whom this ID is assigned, the movement path on the processing target frame image (that is, the movement path in the imaging area of the imaging device 10) can be obtained from the object map, and the object can be tracked.

  The object detection function unit 31 corresponds to the object detection unit referred to in the present invention, and the object identification function unit 32 corresponds to the identification unit referred to in the present invention.

  The movement speed determination function unit 34 determines whether the movement speed of the person being tracked is normal or abnormal by using the determination criterion set by the determination criterion setting function unit 36 described later.

  Note that the abnormal movement speed mentioned here includes not only a person whose movement speed is higher than a certain level compared to a general person but also a person whose movement speed is a certain degree slower than a general person. .

  The tabulation function unit 35 tabulates the amount of change in the position of the person on the processing target frame image between temporally continuous processing target frame images using the above-described object map. This aggregation is performed in a plurality of stages according to the size S of the person on the processing target frame image.

  Even for the same person, the size S on the processing target frame image increases as the distance between the person and the imaging device 10 decreases. Even if the moving distance in the real space is the same, the amount of change on the processing target frame image increases as the distance between the moving location and the imaging device 10 decreases. That is, the totaling function unit 35 calculates the total amount of change in the position of the person on the processing target frame image by classifying the size S of the person on the processing target frame image in a plurality of stages. It is divided according to the distance between people and people.

  The tabulation function unit 35 refers to the object map for each stage divided by the detected human size S, and the person on the processing target frame image between two processing target frame images that are temporally continuous. The distribution of the amount of change in the horizontal direction and the distribution of the amount of change in the vertical direction of the person on the processing target frame image are tabulated. The amount of change in the horizontal direction is an absolute value of the difference between the detection positions in the horizontal direction of a person in two temporally continuous frame images to be processed. Further, the amount of change in the vertical direction is the absolute value of the difference between the detection positions in the vertical direction of a person in two temporally continuous frame images to be processed.

  The determination criterion setting function unit 36, based on the totaling result of the totaling function unit 35, the range of the normal change amount in the horizontal direction on the processing target frame image and the processing target frame image according to the size of the person on the processing target frame image. A criterion for associating the range of the normal change amount in the horizontal direction is set. FIG. 5 is a diagram showing a determination criterion table in which the determination criterion is set by the determination criterion setting function unit 36. In the example of the determination criterion table shown in FIG. 5, the determination criterion for a person whose size S range on the processing target frame image is S0 to S1 is a range where the horizontal change amount is X1 to X2, and This is a setting for determining that the movement speed is normal for a person whose vertical variation is in the range of Y1 to Y2. The same applies to other examples of the range of the size S on the processing target frame image shown in FIG.

  Although not specifically shown, the image processing unit 3 includes an image memory for performing the above-described object detection processing, a memory for storing the object map shown in FIG. 3, and the determination criterion table shown in FIG. Have.

  Next, the operation of the image processing apparatus 1 according to this example will be described. The image processing apparatus 1 according to this example performs the following tracking process, determination criterion setting process, and speed abnormality person detection process. The tracking process is a process of registering the size S and the position (x, y) on the processing target frame image for the person captured in the processing target frame image in the object map shown in FIG. The criterion setting process is a process for setting the criterion table shown in FIG. The abnormal speed person detection process is a process for detecting a person whose movement speed in the imaging area of the imaging apparatus 10 is abnormal.

  FIG. 6 is a flowchart showing the tracking process. The image processing apparatus 1 determines a processing target frame image to be processed this time by the image processing unit 3 (s1). In s1, a processing target frame image extracted based on a predetermined interval from a moving image (moving image captured by the imaging device 10) input to the moving image input unit 2, and is unprocessed at that time The processing target frame image that is the fastest in terms of time (the processing target frame image positioned immediately after the processing target frame image processed immediately before) is determined as the processing target frame image to be processed this time.

  In the object detection function unit 31, the image processing unit 3 detects an object (person) captured in the processing target frame image determined in s1 (s2). In s2, a person imaged by a known background difference method or inter-frame difference method is detected, and a temporary ID is assigned to the detected person. The image processing unit 3 performs identification processing in the object identification function unit 32 to associate the person detected in s2 with the person already detected at this time (s3).

  FIG. 7 is a flowchart showing the identification processing according to s3. The object identification function part 32 extracts the feature-value for every person detected by s2 (s11). In s11, the detected RGB brightness value of the person on the processing target frame image may be extracted as a feature amount, or the shape or color of a face part such as an eye, nose, or mouth may be extracted as a feature amount. Good.

  The object identification function unit 32 includes the person imaged in the previous processing target frame image (that is, the processing target frame image immediately before the current processing target frame image), and the person detected in s2. Are associated (s12). In s12, for each person detected in s2, the feature amount extracted in s11 for that person and the feature amount of each person captured in the previous processing target frame image (this feature amount is determined based on the previous processing target frame). The degree of similarity is calculated. For example, as shown in FIG. 8, person A, person B, and person C are detected in the previous process target frame image, and person a, person b, and person c are detected in the current process target frame image. If detected, the similarity of the feature quantity is calculated for each of the person A, person B, and person C for the person a. Similarly, with respect to the person b, the similarity of the feature quantity is calculated for each of the person A, the person B, and the person C, and the similarity of the feature quantity is also obtained for each of the person A, the person B, and the person C. Calculate the degree.

  The object identification function unit 32 associates each person detected in s2 with a person whose feature value similarity is greater than a predetermined threshold. At this time, if there are a plurality of people whose feature amount similarity is greater than a predetermined threshold, the feature amount is associated with the person with the maximum similarity. If there is no person whose feature quantity similarity is greater than a predetermined threshold, the person is not associated with the person imaged in the previous processing target frame image.

  The person associated with the person imaged in the previous processing target frame image in s12 discards the temporary ID and sets the real ID assigned to the person associated this time. Also, if the real ID is not given to the person captured in the previous processing target frame image associated in s12 (if the temporary ID is held), these temporary IDs are discarded and newly added. Set the generated book ID. Moreover, the person who was not matched with the person imaged by the last process target frame image by s12 hold | maintains temporary ID provided to the person. The person holding the temporary ID is the person who entered the imaging area of the imaging device 10 between the imaging timing of the previous processing target frame image and the imaging timing of the current processing target frame image, or the previous processing target. For example, a person who has not been properly detected by processing the frame image.

  The object identification function unit 32 determines whether the two-stage identification flag is turned on or off (s13). As will be described later, this two-stage identification flag may be used when there is a person who has not been given the real ID (has a temporary ID) among those who have been captured in the previous processing target frame image. This is turned on when there is a person who has not been associated with the person imaged in the previous process target frame image among the persons imaged in the previous process target frame image.

  When the two-stage identification flag is turned on, the object identification function unit 32 is a person who has been imaged in the current processing target frame image and has not been given this ID at this time (the previous one in s12) The person who has not been associated with the person captured in the processing target frame image is associated with the person who has been captured in the previous processing target frame image (s14). In s14, even if the person is captured in the previous processing target frame image, the person associated with the person captured in the previous processing target frame image is not subject to the processing related to s14. It is. In other words, a person who has been captured in the previous processing target frame image and is not associated with a person who has been captured in the previous processing target frame image is the target. Also in the processing related to s14, similar to s12 described above, the similarity of the feature amount is calculated and associated.

  The object identification function unit 32 determines whether there is a person holding a temporary ID among the people imaged in the current processing target frame image, and the information of the person imaged in the previous processing target frame image. It is determined whether there is a person who has not been associated with the person captured in the current processing target frame image (s15, s16).

  The object identification function unit 32 does not have a person who holds the temporary ID among the people imaged in the current processing target frame image and the person who has been imaged in the previous processing target frame image. If there is no person who is not associated with the person imaged in the current processing target frame image, the two-stage identification flag is set to off (s17). In other cases, the two-stage identification flag is turned on. (S18), and the identification process is terminated.

  The above identification process will be described with a specific example. First, the case illustrated in FIG. 9 in which the person C enters the imaging area of the imaging apparatus 10 between the imaging timing of the processing target frame image two times before and the imaging timing of the previous processing target frame image will be described as an example. To do. The person A captured in each processing target frame image shown in FIG. 9 is the same person. Similarly, person B is the same person, and person C is also the same person.

  In FIG. 9, the person C is not associated with any person captured in the previous processing target frame image in the identification processing for the previous processing target frame image. Therefore, at the time when the identification process for the previous processing target frame image is completed, the person C is in a state of holding the temporary ID (a state in which the real ID is not assigned). That is, the two-stage identification flag is set to ON in s18 of the identification process for the previous processing target frame image.

  In s12 of the identification process for the current process target frame image shown in FIG. 9, the person C imaged in the current process target frame image is associated with the person C imaged in the previous process target frame image. At this point, this person C is given a real ID.

  Therefore, the real ID can be given to the person C who has entered the imaging area of the imaging device 10 at an arbitrary timing, and the tracking process can be started appropriately.

  In the example illustrated in FIG. 9, the person A captured in the current processing target frame image is associated with the person A captured in the previous processing target frame image, and is captured in the current processing target frame image. The person B who is being associated is associated with the person B imaged in the previous processing target frame image.

  Next, the person C leaves the imaging area of the imaging apparatus 10 from the imaging area to the outside of the imaging area between the imaging timing of the previous processing target frame image shown in FIG. 10 and the imaging timing of the previous processing target frame image. This will be described as an example. The person A captured in each processing target frame image shown in FIG. 10 is the same person. Similarly, person B is the same person, and person C is also the same person.

  In FIG. 10, the person C captured in the previous processing target frame image is not associated with the person captured in the previous processing target frame image in the identification processing for the previous processing target frame image. For this reason, the two-stage identification flag is set to ON in s18 of the previous identification process. Then, in s12 of the identification process for the current processing target frame image shown in FIG. 10, the person A captured in the current processing target frame image is associated with the person A captured in the previous processing target frame image. The person B captured in the current processing target frame image is associated with the person B captured in the previous processing target frame image. On the other hand, in the identification process s14, the person C captured in the previous processing target frame image is not associated with the person captured in the current processing target frame image.

  Therefore, the tracking process for the person C who has left the imaging area of the imaging device 10 out of the imaging area can be properly terminated.

  On the other hand, in the description of FIG. 10 described above, the person C who has left the imaging area of the imaging device 10 does not actually leave the imaging area, and as shown in FIG. In the object detection processing for the target frame image, if the person C is not properly detected for some reason, the person C is captured in the current processing target frame image, and thus is identified for the current processing target frame image. In s12 of the process, it is not associated with any person imaged in the previous process target frame image. However, this person C is associated with the person C who has been captured in the previous processing target frame image in s14 of the identification processing for the current processing target frame image. Therefore, it is possible to properly continue the tracking process for a person who has lost his sight temporarily (a person who cannot be properly detected for some reason in the object detection process of s2).

  Further, as shown in FIG. 12, since the person A and the person B overlap in the previous process target frame image, the identification process for the previous process target frame image overlaps the previous process target frame image. A case will be described in which the person A and the person B are not associated with the person A and the person B captured in the previous processing target frame image. In this case, the two-stage identification flag is set to ON in s18 of the previous identification process.

  In the example shown in FIG. 12, the person A captured in the current processing target frame image corresponds to the person A captured in the previous processing target frame image in s14 of the identification process for the current processing target frame image. The person B that is attached and is captured in the current processing target frame image is associated with the person B that is captured in the previous processing target frame image.

  Therefore, even if a person overlaps on the processing target frame image, for example, a situation in which it is impossible to properly associate with the person captured in the previous processing target frame image, the tracking process for those persons is performed appropriately. Can continue.

  The three processing target frame images shown in FIGS. 9 to 12 are three processing target frame images that are temporally continuous.

  Further, in the above example, the detected person is associated with each other between the processing target frame images using the detected person's feature amount, but may be performed using the moving direction or the like.

  When the object map creation function unit 33 of the image processing unit 3 completes the identification processing related to s3, the size S on the processing target frame image and the processing target are displayed on the object map shown in FIG. The position of the frame image is registered (s4), and the process returns to s1.

  As described above, in the image processing apparatus 1 according to this example, the identification process shown in FIG. 7, particularly the process according to s <b> 14, is executed, thereby temporarily losing an object (in this example, a person) or a processing target. Objects that overlap on the frame image can also be properly tracked.

  Next, determination criterion setting processing of the image processing apparatus 1 according to this example will be described. This determination criterion setting process is executed every day at an appropriate timing such as midnight. FIG. 13 is a flowchart showing determination criterion setting processing.

  The counting function unit 35 refers to the object map created by the object map creation function unit 33, and performs two processes that are continuous in time for each stage in which the detected human size S on the processing target frame image is divided. The distribution of the variation in the horizontal direction on the processing target frame image of the detected person between the target frame images and the distribution of the variation in the vertical direction on the processing target frame image of the detected person are tabulated (s21). FIG. 14 is a diagram illustrating a totaling result of totaling the amount of change in the horizontal direction for a certain size S. The horizontal axis in FIG. 14 is the amount of change in the horizontal direction, and the vertical axis is the frequency (the number of extractions extracted from the object map).

  The determination criterion setting function unit 36, based on the counting result of the counting function unit 35, has two processing target frame images that are temporally continuous at each stage where the detected human size S on the processing target frame image is divided. The normal range of the amount of change in the horizontal direction on the processing target frame image of the detected person and the normal range of the amount of change in the vertical direction on the processing target frame image of the detected person are determined (s22). As shown in FIG. 14, the normal range determined in s22 is determined as a normal range including a change amount of 60% to 70% with the peak at approximately the center.

  The determination criterion setting function unit 36 processes the detected person between two processing target frame images that are temporally continuous at each stage in which the size S of the detected person on the processing target frame image is divided in s22. When the normal range of the amount of change in the horizontal direction on the target frame image and the normal range of the amount of change in the vertical direction on the detected processing target frame image of the person are determined, these are set in the determination criterion table shown in FIG. (S23), and this process ends.

In this example, the amount of change in the horizontal direction and the amount of change in the vertical direction are individually set in the determination criterion table. However, the position of the detected person between two temporally consecutive frame images to be processed is determined. The amount of change
((Horizontal change amount) ² + (Vertical change amount) ² ) ^1/2
It may be configured to set.

  Next, speed abnormality person detection processing will be described. FIG. 15 is a flowchart showing a speed abnormality person detection process. This speed abnormality person detection process is performed every time the object map creation function unit 33 executes s4 in the tracking process described above. Specifically, the moving speed determination function unit 34 determines that the object map creation function unit 33 performs the tracking process s4 described above, and the object map shown in FIG. This speed abnormality person detection process is executed for a person who registers a position on the image (hereinafter referred to as a determination target person).

  The moving speed determination function unit 34 determines the normal range of the horizontal change amount and the vertical change amount according to the size S on the processing target frame image of the determination target person from the determination reference table shown in FIG. The normal range is read (s31).

  The moving speed determination function unit 34 refers to the object map for the determination target person, and calculates the horizontal direction change amount on the processing target frame image and the vertical direction change amount on the processing target frame image (s32). In s32, the position on the two latest processing target frame images is used for this determination target person.

  The moving speed determination function unit 34 determines whether the amount of change in the horizontal direction on the processing target frame image calculated in s32 is in the normal range, and the amount of change in the vertical direction on the processing target frame image calculated in s32 is in the normal range. It is determined whether there is (s33, s34). In the moving speed determination function unit 34, the amount of change in the horizontal direction on the processing target frame image calculated in s32 is in the normal range, and the amount of change in the vertical direction on the processing target frame image calculated in s32 is in the normal range. If there is, it is determined that the speed is not abnormal (s35). On the other hand, if at least one of the horizontal change amount on the processing target frame image or the vertical change amount on the processing target frame image calculated in s32 is not within the normal range, the moving speed determination function unit 34 determines that the speed is abnormal. (S36).

  FIG. 15 shows a process for determining whether or not a single determination target person is a speed abnormality person. When there are a plurality of determination target persons, FIG. Repeat the process shown.

  When the movement speed determination function unit 34 determines that the person to be determined is a speed abnormality person in s36, the image processing apparatus 1 notifies the higher-level apparatus to that effect from the output unit 4.

  As described above, the image processing apparatus 1 can calculate the speed of a person detected by processing the processing target frame image extracted from the moving image captured by the imaging apparatus 10 without calculating the position of the person in the real space. Determine if you are an abnormal person. Therefore, the processing load of the image processing apparatus can be greatly reduced, and the cost of the apparatus main body can be reduced. Further, when installing the imaging apparatus 10, it is not necessary to set parameters for associating the position on the processing target frame image to be imaged with the position in the real space, and the installation operation of the imaging apparatus 10 can be easily performed.

  Further, the image processing apparatus 1 determines whether or not the person is an abnormal speed by using the determination criterion set in the determination criterion table shown in FIG. 5 by the determination criterion setting process shown in FIG. Therefore, according to the environment of the imaging area of the imaging device 10, it can be determined appropriately whether it is a speed abnormal person.

  Further, in the above example, when the moving speed determination function unit 34 determines that the determination target person is a speed abnormality person in s36, the image processing apparatus 1 notifies the higher-level apparatus to that effect from the output unit 4. In response to a request from the terminal device, a moving image of a portion where the determination target person who has been determined to be a speed abnormality person in s36 is captured may be transmitted to the terminal apparatus. In the moving image input to the moving image input unit 2, the part where the determination target person determined to be a speed abnormality person in s 36 is captured is obtained from the object map shown in FIG. 3. Specifically, by searching the object map using the main ID of the determination target person determined to be a speed abnormality person in s36, the processing target from which the determination target person related to this main ID was first imaged It is possible to identify the frame image and the processing target frame image that was captured last. The image processing apparatus 1 performs processing immediately after the processing target frame image captured last from the processing frame image immediately before the processing target frame image captured first by the determination target person determined to be a speed abnormality person. The moving image part up to the target frame image (all the frame images located in the meantime) is transmitted to the terminal device.

  Further, the image processing apparatus 1 transmits, to the terminal device, a frame image in which the determination target person (speed abnormal person) is captured for each determination target person determined to be a speed abnormal person in s36. Alternatively, it may be possible to easily select a speed abnormal person who requests the image processing apparatus 1 to transmit a moving image. Specifically, when the operator selects any one of the frame images in which the speed abnormality person is captured in the terminal device, the moving image in which the speed abnormality person is captured in the selected frame image of the terminal device. The image processing apparatus 1 may be requested to transmit an image.

  In the above example, the object to be detected is a person, but other types of moving bodies such as vehicles may be used.

1-Image processing device 2-Moving image input unit 3-Image processing unit 4-Output unit 10-Imaging device 31-Object detection function unit 32-Object identification function unit 33-Object map creation function unit 34-Movement speed determination function unit 35-Aggregating function unit 36-Criteria setting function unit

Claims (5)

A moving image input unit to which a moving image in which the imaging device captures the target area is input;
A frame image extracted at a predetermined interval from the moving image input to the moving image input unit is processed as a processing target frame image, an object captured in the processing target frame image is detected, and the detected object An object detection unit for detecting the position of
An identification unit that associates the object detected by the object detection unit between processing target frame images,
The identification unit is an image processing apparatus that uses three processing target frame images that are temporally continuous and associates an object captured between the three processing target frame images.   Based on the similarity of the imaged object between two processing target frame images that are not temporally intermediate among the three processing target frame images that are temporally continuous, The image processing apparatus according to claim 1, wherein an object captured between two processing target frame images is associated.   The identification unit is not positioned in the middle in time when a plurality of objects are overlapped on the processing target frame image in the middle in time among the three processing target frame images in time. The image processing apparatus according to claim 1, wherein objects captured between these three processing target frame images are associated with each other based on the similarity of the captured object between the two processing target frame images. An object captured in the processing target frame image by processing a frame image extracted at a predetermined interval from a moving image in which the target area is captured by the imaging device and input to the moving image input unit. And an object detection step for detecting the position of the detected object,
An identification step of associating the object detected in the object detection step between the processing target frame images,
The tracking method, wherein the identifying step uses the three processing target frame images that are temporally continuous, and associates the imaged object between the three processing target frame images. An object captured in the processing target frame image by processing a frame image extracted at a predetermined interval from a moving image in which the target area is captured by the imaging device and input to the moving image input unit. And an object detection step for detecting the position of the detected object,
An identification step for associating the object detected in the object detection step between the processing target frame images, and a tracking program for causing a computer to execute the identification step,
In the tracking program, the identification step uses the three processing target frame images that are temporally continuous, and associates an object captured between the three processing target frame images.

Images