JP2016152467A - Tracking device, tracking method and tracking program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a tacking device capable of tracking an object accurately.SOLUTION: A tracking device 10 includes: an object person DB22 for storing the representative color data indicating the representative color for tracking a tracking object; a moving object detection processing section 12 which generates a first foreground background separation data for separating the foreground data and background data based on an acquired video data; a mask processing section 13 to data outside object person, which generates second foreground background separation data based on a region of a color including the representative color out of the region of the video data corresponding to the first foreground background separation data; and a tracking processing section 16 for tracking the tracking object based on at least one of the first foreground background separation data and second foreground background separation data.SELECTED DRAWING: Figure 1

Description

  The present invention relates to a tracking device, a tracking method, and a tracking program, and more particularly, to a tracking device, a tracking method, and a tracking program that use a representative color for tracking an object.

  2. Description of the Related Art A technique is known in which an object such as a person is imaged by an imaging device, and the object is automatically tracked using the captured video data.

  In the prior art, color information serving as a reference for an object is stored in advance, and tracking is performed by detecting the object from video data based on the color information. In such a conventional technique, when the color of an object changes due to the influence of illuminance or the like, the object cannot be detected and tracking may fail.

  Therefore, for example, in Patent Document 1, the monitor indicates that the tracking has been performed correctly by displaying on the screen that the color of the object has changed.

JP 2010-226644 A

  However, in the case where a plurality of objects are included in the video data, there is a high possibility that occlusion occurs when the object to be tracked overlaps with another object. However, in the conventional technology such as Patent Document 1, tracking processing at the time of occurrence of occlusion is performed. Not considered.

  Therefore, in the prior art such as Patent Document 1, there is a problem that when an occlusion occurs, an object cannot be accurately tracked.

  Accordingly, the present invention provides a representative color storage unit that stores representative color data indicating a representative color for tracking a tracking target object, and a first foreground that separates foreground data and background data based on the acquired video data. A first foreground / background separation data generation unit for generating background separation data; and a second foreground based on a color region including the representative color among the regions of the video data corresponding to the first foreground / background separation data. A second foreground / background separation data generation unit that generates background separation data, and a tracking processing unit that tracks the tracking object based on at least one of the first foreground / background separation data or the second foreground / background separation data And a tracking device comprising:

  In addition, the present invention stores representative color data indicating a representative color for tracking a tracking object, and generates first foreground / background separation data that separates foreground data and background data based on the acquired video data. Then, second foreground / background separation data is generated based on a color region including the representative color among the regions of the video data corresponding to the first foreground / background separation data, and the first foreground / background separation data Alternatively, a tracking method is provided in which the tracking target is tracked based on at least one of the second foreground / background separation data.

  Furthermore, the present invention stores representative color data indicating a representative color for tracking a tracking object, and generates first foreground / background separation data for separating foreground data and background data based on the acquired video data Then, second foreground / background separation data is generated based on a color region including the representative color among the regions of the video data corresponding to the first foreground / background separation data, and the first foreground / background separation data Alternatively, a tracking program for tracking the tracking object based on at least one of the second foreground / background separation data and causing a computer to execute a tracking process is provided.

  According to the present invention, it is possible to provide a tracking device, a tracking method, and a tracking program that can accurately track an object.

1 is a configuration diagram showing a configuration of a tracking system according to Embodiment 1. FIG. FIG. 6 is an explanatory diagram for explaining an outline of an operation of the apparatus according to the first embodiment. FIG. 6 is an explanatory diagram for explaining an outline of an operation of the apparatus according to the first embodiment. FIG. 6 is an explanatory diagram for explaining an outline of an operation of the apparatus according to the first embodiment. FIG. 6 is an explanatory diagram for explaining an outline of an operation of the apparatus according to the first embodiment. FIG. 6 is an explanatory diagram for explaining an outline of an operation of the apparatus according to the first embodiment. 3 is a flowchart showing the operation of the apparatus according to the first embodiment. 3 is a flowchart showing the operation of the apparatus according to the first embodiment. It is a figure which shows the basic technique of a tracking method. It is a figure which shows the basic technique of a tracking method.

  First, before describing the embodiment, a basic technique for automatically tracking an object will be described.

  As shown in FIG. 8, in the basic technique, representative color data 911 indicating the representative color that is a feature of the subject 901 is acquired for each frame image of the video data 900 and stored in the representative color data storage unit 910. An object person 901 in the frame image is detected based on the color data 911. Further, in the basic technique, the representative color data 911 is periodically updated, so that the subject 901 can be tracked even when the color of the subject 901 changes due to the influence of illuminance or the like.

  In such a basic technology, foreground / background separation data for separating the subject's area is generated by performing motion detection on the video data, and the representative color of the subject is extracted using the foreground / background separation data. To do.

  For example, as illustrated in FIG. 9, when a plurality of persons overlap in a frame image 920 and occlusion occurs, when foreground detection is performed on the frame image 920, the foreground / background separation data 921 includes regions of the plurality of persons. It becomes an image. Then, since the representative color extraction image 922 obtained by combining the frame image 920 and the foreground / background separation data 921 includes images of a plurality of persons, the representative color data extracted from this image includes the representative colors of the plurality of persons. Therefore, it is impossible to extract the representative color of only the target person.

  As a result, in the basic technology, when occlusion occurs, tracking is performed based on representative color data including representative colors of persons other than the target person, so that only the desired target person can be accurately tracked. Have difficulty. In addition, if the target person is tracked using multiple cameras and occlusion occurs from the beginning when the camera is switched, the target person and the non-target data cannot be separated. In this case, only the target person is tracked. It is not possible.

  Therefore, in the following embodiment, even when occlusion occurs, the representative color of the target object is reliably extracted, and the target person can be accurately tracked.

(Embodiment 1)
Hereinafter, embodiments will be described with reference to the drawings. FIG. 1 shows a configuration of a tracking system 1 according to the present embodiment.

  As shown in FIG. 1, the tracking system 1 includes a tracking device 10, a storage device 20, and an input device 30. Note that each device may be one device or any number of devices instead of separate devices. For example, a single tracking device including the configuration of the tracking device 10 and the storage device 20 may be used.

  The input device 30 is a device that inputs information necessary for the processing of the tracking device 10. For example, the input device 30 includes a camera 31 and a mouse 32.

  The camera 31 is an imaging device that incorporates an imaging sensor, and uses this imaging sensor to capture an image of the subject 300 that is the object. The camera 31 generates video data including a plurality of captured frame images and outputs the generated video data to the tracking device 10. Note that video data may be generated by a device other than the camera 31.

  The mouse 32 is a pointing device for a user to operate a GUI (Graphical User Interface) displayed on a video display unit (not shown) of the tracking device 10. When the user performs a click operation according to the display on the video display unit, the mouse 32 outputs the coordinate data of the clicked position to the tracking device 10. The display on the video display unit may be operated by a device other than the mouse 32.

  The storage device 20 is a device that stores information necessary for the processing of the tracking device 10. For example, the storage device 20 includes a video data DB (database) 21 for video and image data, a subject DB 22 for subject data to be tracked, and a representative color learning DB 23 for learning a representative color of the subject. I have.

  The video data DB 21 includes video data 211 captured from the target person 300, background data 212 that is the background of the video data 211, first foreground / background separation data 213 for separating the foreground and background from the video data 211, and second data. Foreground / background separation data 214 and the like are stored. The target person DB 22 stores feature data 221 indicating the characteristics of the target person 300, and the feature data 221 includes representative color data 222 indicating the representative color of the target person 300, shape data 223 indicating the shape of the target person 300, and the like. It is. For example, it can be said that the subject DB 22 is a representative color storage unit that stores the representative color data 222. The representative color learning DB 23 stores representative color boundary range data 231 indicating the boundary range of the representative color of the subject 300 and the like.

  The tracking device 10 is an image processing device that analyzes input video data and tracks the subject 300. The tracking device 10 includes an image analysis processing unit 11, a moving object detection processing unit 12, a target person data outside mask processing unit 13, a designated target person data extraction unit 14, a representative color learning unit 15, and a tracking processing unit 16.

  In addition, as long as the operation | movement which concerns on this Embodiment mentioned later is realizable, you may comprise by another functional block. Each function (each process) in the tracking device in FIG. 1 is configured by hardware and / or software, and may be configured by one piece of hardware or software, or may be configured by a plurality of pieces of hardware or software. Also good. Each function of the tracking device may be realized by a computer having a CPU (Central Processing Unit), a memory, and the like. For example, a program for performing an operation according to the present embodiment to be described later may be stored in the storage device 20, and each function of the tracking device may be realized by executing the program stored in the storage device 20 by the CPU. Good.

  The image analysis processing unit 11 is a video acquisition unit that acquires video data 211 captured by the camera 31 of the target person 300, and performs image conversion processing on the acquired video data 211 as necessary. The image analysis processing unit 11 performs image conversion processing such as resizing processing and color conversion processing on the acquired video data 211, outputs the video data 211 after the image conversion to the moving object detection processing unit 12, and the video data DB 21 To remember. Note that this image conversion process may be arbitrarily executed as necessary, and may be omitted in the following description.

  The moving object detection processing unit 12 generates background data 212 from the video data 211 acquired and converted by the image analysis processing unit 11, stores the generated background data 212 in the video data DB 21, and is a target data outside mask processing unit. 13 and the designated target person data extraction unit 14. The background data 212 is image data of only the background not including the foreground portion moving from the video data 211.

  In addition, the moving object detection processing unit 12 generates a first foreground / background separation data 213 by executing a moving object detection process on the video data 211 based on the background data 212, and generates the generated first foreground / background separation data 213. Is stored in the video data DB 21 and is output to the mask data processing unit 13 and the designated target data extraction unit 14. It can be said that the moving object detection processing unit 12 is a first foreground / background separation data generation unit that generates first foreground / background separation data. The first foreground / background separation data 213 is image data for separating the foreground and the background from the video data 211, and is image data obtained by extracting only a moving part from the video data 211 by moving object detection. The first foreground / background separation data may be generated based on the moving object detection processing result for the video data 211, or the first foreground / background separation data may be generated based on the difference between the video data 211 and the background data 212. May be. By using the difference between the moving object detection process and the background data, it is possible to reliably generate the mask including the moving subject.

  The designation target person data extraction unit 14 determines the characteristics of the target person 300 from the background data 212 and the first foreground / background separation data 213 generated by the moving object detection processing unit 12 according to the designation of the coordinate data input from the mouse 32. Data 221 is extracted, and the extracted feature data 221 is stored in the target person DB 22 and output to the representative color learning unit 15. The designated target person data extraction unit 14 performs an object mapping process for human detection (object detection) on an area including coordinates specified by the user in the video data 211, so that the contour shape of the target person 300 is obtained. Is extracted to generate the shape data 223, and the representative color in the contour shape is extracted to generate the representative color data 222. The representative color of the target person 300 is a color that represents the target person 300 and is a characteristic color indicating a feature for identifying the target person 300.

  The representative color learning unit 15 sets a representative color boundary range (representative color range) using the representative color data 222 of the target person extracted by the designated target person data extraction unit 14, and the set representative color boundary range Data (representative color range data) 231 is stored in the representative color learning DB 23. The representative color learning unit 15 can be said to be a representative color update unit that updates the representative color data with the representative color of the area of the video data 211 corresponding to the second foreground / background separation data 214 as the representative color of the tracking target object 300. .

  The target-data-out mask processing unit 13 uses the representative color boundary range data 231 stored in the representative color learning DB 23, the background data 212 generated by the moving object detection processing unit 12, and the first foreground / background separation data 213. Then, a mask process for masking image data other than the target person 300 is performed, and the background data 212 after the mask process is overwritten on the video data DB 21.

  The subject data out-of-object mask processing unit 13 determines whether the representative color boundary range data 231 includes the color of the portion corresponding to the foreground of the first foreground / background separation data 213 in the video data 211. The foreground / background separation data 214 is generated, and the generated second foreground / background separation data 214 is stored in the video data DB 21. The subject-data-outside mask processing unit 13 generates second foreground / background separation data 214 based on the color region including the representative color in the video data 211 corresponding to the first foreground / background separation data 213. It can also be said to be a foreground / background separation data generation unit. The second foreground / background separation data 214 is image data for separating the foreground and the background from the video data 211, and is image data obtained by extracting only a portion corresponding to the representative color from the video data 211. The binarized data may be used as the second foreground / background separation data 214 depending on whether the color of the corresponding part matches the representative color or whether the color of the corresponding part is within the range of the representative color. Data binarized according to whether or not may be used as the second foreground / background separation data 214. By using the coincidence with the representative color as a reference, a mask for at least the subject can be generated. By using the range of the representative color as a reference, a mask for only the target person can be generated more appropriately based on the learning history.

  The tracking processing unit 16 performs processing for tracking the target person 300 in the video data 211 based on the feature data 221 of the target person 300. The tracking processing unit 16 recognizes the target person 300 from the area closest to the representative color data 222 and the shape data 223 of the target person 300, and automatically tracks the target person 300. Furthermore, the tracking processing unit 16 can perform the tracking process using the first foreground / background separation data 213 and the second foreground / background separation data 214. Since the tracking process based on the feature data 221 is performed on the foreground part extracted from the first foreground / background separation data 213, only the moving part can be processed, so that the tracking process can be performed efficiently. Further, by performing the tracking process based on the feature data 221 on the foreground part extracted from the second foreground / background separation data 214, only the part corresponding to the representative color can be processed, so that the tracking process can be performed more efficiently. It can be carried out.

  The outline | summary of operation | movement of the tracking apparatus 10 which concerns on this Embodiment is demonstrated using FIGS. First, as illustrated in FIG. 2, the moving object detection processing unit 12 uses the motion vector obtained by the moving object detection process for the frame image (color image) 211 a included in the acquired video data 211 to detect the target person 300. A process for separating the foreground and the background is performed, and first foreground / background separation data 213 is generated (step S10).

  Further, the designation target person data extraction unit 14 combines the frame image 211a and the first foreground / background separation data 213 to cut out the image of the foreground part to generate a representative color extraction image 215. The designated target person data extraction unit 14 extracts a representative color (feature amount) serving as a reference for the target person 300 from the representative color extraction image 215, and generates representative color data 222 (step S11). Here, representative colors of the head, torso, and legs are acquired for each part of the subject 300, and representative color data of the head, representative color data of the torso, and representative color data of the legs are generated. Further, representative color data of the front, back, and side of the target person 300 (each part) may be generated.

  Subsequently, since unevenness occurs in the brightness of the image outdoors and indoors, and it is necessary to update the feature amount appropriately for the tracking process, as shown in FIG. The representative color data 222 of the target person 300 is periodically updated (step S12). For example, the designation target person data extraction unit 14 uses the first foreground / background separation data 213 for the frame image 211a of the 0th frame, the 25th frame, and the 50th frame of the video data 211 to use the representative color extraction image 215. The representative color of each part of the subject is extracted, and the representative color data 222 is updated.

  Further, the representative color learning unit 15 learns the boundary range of the representative color of the target person 300 between the specific frame images, and generates the representative color boundary range data 231 (step S13). For example, the representative color data 222 of the head in each frame changes in the order of R (red) = 253, 239, 247, changes in the order of G (green) = 228, 218, 167, and B (blue) = 208. , 143, 102, the head representative color boundary range data 231 includes an R range of 239 ≦ R ≦ 253 and a G range of 167 ≦ G ≦ 228 so as to include these color changes. , B is set as 102 ≦ B ≦ 208. The subsequent processing may be performed using only the representative color without learning the boundary range of the representative color.

  Subsequently, after creating (learning) the representative color boundary range data 231, as shown in FIG. 4, the subject data outside mask processing unit 13 generates the first foreground / background separation data generated using the motion vector. 213 and the frame image 211a are combined to generate a representative color extraction image 215. Further, the representative color extraction image 215 is compared with the representative color boundary range data 231 of the subject 300, so that the second foreground / background separation is performed. Data 214 is created (step S14). At this time, if the color of the representative color extraction image 215 is within the boundary range of the representative color, the pixel of the second foreground / background separation data 214 is set to white, otherwise it is set to black and binarized. The foreground / background separation data 214 is generated. Further, for each part of the subject person, it is determined whether it is within the boundary range of the representative color of the corresponding part.

  Subsequently, as shown in FIG. 5, as in FIG. 2, the designated target person data extraction unit 14 generates a representative color extraction image 215 from the frame image 211 a and extracts the representative color data 222 of the target person 300 ( Step S15). After that, as in FIG. 3, the representative color learning unit 15 further learns (updates) the boundary range of the representative colors.

  By using such a method, only a portion close to the representative color data of the subject can be extracted as the foreground. As shown in FIG. 6, when the subject person 300 overlaps another person 301 in the frame image 211 a and occlusion occurs, the foreground portion of the first foreground / background separation data 213 using motion detection includes the subject person 300 and The entire other person 301 is included. However, since the foreground portion of the second foreground / background separation data 214 using the representative color is substantially only the target person 300, the representative color extraction image 215 using the second foreground / background separation data 214 is almost identical. A representative color of only the target person 300 can be extracted. That is, foreground / background separation data with higher accuracy than the foreground / background separation data obtained by moving object detection can be created, so that it is possible to prevent unnecessary person data from being included in the representative color data of the subject.

  Next, a detailed operation example of the tracking device 10 according to the present embodiment will be described with reference to FIGS. 7A and 7B. As shown in FIGS. 7A and 7B, first, the image analysis processing unit 11 acquires video data 211 from the camera 31 (step S101). Note that the image analysis processing unit 11 is not limited to the camera 31 and may acquire video data 211 generated in advance from a recording medium or the like.

  Subsequently, the moving object detection processing unit 12 performs a moving object detection process on the acquired video data 211, and generates first foreground / background separation data 213 that separates the background and the foreground using the obtained motion vector. (Step S102). The first foreground / background separation data 213 is binary image data composed of white foreground pixels and black background pixels. For example, black has a pixel value = 0 (minimum value) and white has a pixel value = 255 (maximum value). Note that the moving object detection processing unit 12 may generate the first foreground / background separation data 213 using not only the motion vector but also color information of a difference from a predetermined background image.

  Subsequently, the designation target person data extraction unit 14 extracts a foreground pixel block from the generated first foreground / background separation data 213 (step S103). A foreground pixel block is a group of pixels in which foreground pixels are continuous, and is a collection of pixels that have a continuous white value.

  Subsequently, the designated target person data extraction unit 14 refers to the target person DB 22 and checks whether or not there is feature data 221 including the representative color data 222 of the target person 300 to be tracked (step S104). In step S <b> 104, when the feature data 221 of the subject person 300 is not stored in the subject person DB 22, the designated subject person data extraction unit 14 confirms whether or not the feature data 221 of the subject person 300 can be extracted from the video data 211. (Step S105). Specifically, based on the coordinate data specified by the user with the mouse 32 and the foreground pixel block of the first foreground / background separation data 213, it is confirmed whether or not the specified coordinate data is included in the foreground pixel block. Thus, it is determined whether or not the feature data 221 can be extracted.

  In step S105, when the feature data 221 of the target person 300 cannot be extracted, the process returns to step S101 and the subsequent processing is repeated. On the other hand, if the feature data 221 of the target person 300 can be extracted in step S105, the designated target person data extraction unit 14 uses the foreground pixel block of the first foreground / background separation data 213 and the frame image (color image) of the video data 211. A representative color extraction image 215 is generated by combining the pixels in the designated area 211a, and representative color data 222, which is a representative color of pixel data (color data) of only the foreground portion, is created (step S106). When the creation of the representative color data 222 is completed, the boundary range of the representative color is learned in step S118 and later.

  On the other hand, when the feature data 221 of the subject 300 is stored in the subject DB 22 in step S104, a loop that repeats the processing of steps S108 to S113 by the number of foreground pixel blocks of the first foreground / background separation data 213. Processing is executed (step S107). Here, first, the subject data out-of-object mask processing unit 13 synthesizes the foreground pixels and the frame image (color image) 211a only within the foreground pixel block region of the first foreground / background separation data 213 (step S108). As a result, a feature amount extraction image (representative color extraction image 215) from which a feature amount (representative color) can be extracted only by the foreground portion is generated.

  Subsequently, a loop process is performed in which the processes of steps S110 to S113 are repeated by the number of pixels in the foreground pixel block area of the first foreground / background separation data 213 (S109). Here, first, the subject data outside mask processing unit 13 compares the frame image (color image) 211a corresponding to the foreground pixel block of the first foreground / background separation data 213 and the representative color boundary range data 231 of the subject. (Step S110). At this time, the non-subject data mask processing unit 13 checks whether or not the color of the pixel of the frame image 211a to be compared is within the range of the representative color boundary range data 231 (step S111).

  In step S111, when the color of the pixel of the frame image 211a to be compared is outside the range of the representative color boundary range data 231, the subject-data-outside mask processing unit 13 applies the corresponding pixel of the second foreground / background separation data 214 to be generated. The value is set to 0 (black) (step S112). On the other hand, in step S111, when the color of the pixel of the frame image 211a to be compared is within the range of the representative color boundary range data 231, the subject-data-outside mask processing unit 13 matches the corresponding second foreground / background separation data 214 to be generated. The pixel value to be set is set to 255 (white) (step S113).

  Subsequently, it is determined whether or not the setting of all foreground pixel blocks has been completed (step S114), and if there is an unset foreground pixel block, the process returns to step S107, and the setting of all foreground pixel blocks is completed. Then, the process proceeds to step S115. When there are a plurality of foreground pixel blocks of the first foreground / background separation data 213, Steps S110 to S113 are executed in each foreground pixel block region, and second foreground / background separation data 214 after the representative color processing is generated. The Subsequently, the subject-data-outside mask processing unit 13 performs noise clear processing such as outside light by illumination on the generated second foreground / background separation data 214 (step S115). The second foreground / background separation data 214 is image data obtained by binarizing a background pixel and a foreground pixel using a motion vector and a representative color.

  Subsequently, the tracking processing unit 16 uses the generated second foreground / background separation data 214 and the characteristic data 221 of the target person 300 to execute a tracking process for tracking the target person 300 with respect to the frame image 211a ( Step S116). The tracking processing unit 16 uses the second foreground / background separation data 214 to obtain a centroid value that is the position with the highest likelihood in the feature data 221 of the target person 300 in the frame image 211a, and uses the centroid value as a target. The person 300 is detected. Note that the tracking process may be performed using the first foreground / background separation data 213 obtained by moving object detection. Subsequently, the tracking processing unit 16 extracts a feature amount including the representative color data 222 from the pixel at the position having the highest likelihood in the calculated feature data 221 of the target person 300 (step S117).

  Subsequently, the representative color learning unit 15 learns representative color data using the representative color data 222 of the target person 300 extracted in step S106 or step S117 (step S118). That is, the representative color learning unit 15 sets the boundary range of the representative color so as to include the extracted representative color data 222 of the target person 300, and generates the representative color boundary range data 231.

  Subsequently, the representative color learning unit 15 checks whether or not there is unused representative color data 222 that is not used for a certain period in each part of the target person 300 in the learned representative color data 222 (step S119). For example, within the range of the representative color boundary range data 231, a color that does not correspond in the tracking process for a certain period is determined to be an unused color.

  If there is unused representative color data 222 in step S119, the representative color learning unit 15 deletes the corresponding representative color data from the learned representative color data 222 (step S120). In other words, the representative color boundary range data 231 is set so that the unused representative color data 222 is not included in the range. On the other hand, if there is no unused representative color data in step S119, or after deleting the unused representative color data in step S120, the representative color learning unit 15 determines whether there is new representative color data 222 that can be added. (Step S121).

  In step S121, if there is representative color data 222 to be added, the representative color learning unit 15 registers the newly extracted representative color data 222 in the representative color learning DB 23, sets the boundary range of the representative color, and sets the representative color. The boundary range data 231 is updated (step S122). On the other hand, if there is no representative color data 222 to be added in step S121, or after adding the representative color data 222 in step S122, the process returns to step S101 and the subsequent processing is repeated.

  As described above, in this embodiment, the representative color of the target person to be tracked is learned, and the foreground and the background are separated based on the learned representative color or its range. That is, in the present embodiment, first foreground / background separation data is generated by moving object detection of video data, and further, second foreground / background separation data is generated based on the learned representative color of the subject, The representative color of the subject is extracted using the foreground / background separation data of 2.

  Thereby, only the foreground portion of only the target person can be separated from the video data, and the representative color of only the target person can be extracted. In other words, even if occlusion occurs, it is possible to suppress the extraction of unnecessary data as the feature amount of the target person, so that only the target person can be reliably tracked. In addition, when imaging a target person with multiple cameras, even if occlusion occurs during tracking across cameras, it is possible to separate the target person from other persons and specify only the target person, Even in this case, the target person can be tracked.

  Further, by using the difference between the first foreground / background separation data based on the motion detection and the second foreground / background separation data using the representative color, the ratio of occurrence of occlusion, that is, the percentage of the subject's body. It can be determined whether it is hidden.

  Note that the present invention is not limited to the above-described embodiment, and can be changed as appropriate without departing from the spirit of the present invention.

DESCRIPTION OF SYMBOLS 1 Tracking system 10 Tracking apparatus 11 Image analysis process part 12 Moving body detection process part 13 Target person data outside mask process part 14 Designation target person data extraction part 15 Representative color learning part 16 Tracking process part 20 Storage device 21 Video data DB
22 Target DB
23 representative color learning 30 input device 31 camera 32 mouse 211 video data 211a frame image 212 background data 213 first foreground / background separation data 214 second foreground / background separation data 215 representative color extraction image 221 feature data 222 representative color data 223 Shape data 231 Representative color boundary range data 300 Target person 301 Other person

Claims (7)

A representative color storage unit that stores representative color data indicating a representative color for tracking a tracking object;
A first foreground / background separation data generation unit that generates first foreground / background separation data that separates foreground data and background data based on the acquired video data;
A second foreground / background separation data generation unit that generates second foreground / background separation data based on a color region including the representative color in the region of the video data corresponding to the first foreground / background separation data;
A tracking processing unit that tracks the tracking object based on at least one of the first foreground / background separation data or the second foreground / background separation data;
A tracking device comprising: The first foreground / background separation data generation unit generates the first foreground / background separation data based on a moving object detection processing result for the video data or a difference between the video data and the background data.
The tracking device according to claim 1. The representative color data includes a range of representative colors of the tracking object,
The second foreground / background separation data generation unit determines whether or not the color of the area of the video data corresponding to the first foreground / background separation data matches the representative color or within the range of the representative color. The data binarized according to the above is used as the second foreground / background separation data,
The tracking device according to claim 1. A representative color update unit that updates the representative color data so that a representative color of the region of the video data corresponding to the second foreground / background separation data is included in the range of the representative color;
The tracking device according to any one of claims 1 to 3. The tracking processing unit tracks the tracking object in the video data based on at least one of the first foreground / background separation data or the second foreground / background separation data and the representative color data;
The tracking device according to any one of claims 1 to 4. Storing representative color data indicating a representative color for tracking a tracking object;
Generating first foreground / background separation data for separating the foreground data and the background data based on the acquired video data;
Generating second foreground / background separation data based on a region of the color including the representative color in the region of the video data corresponding to the first foreground / background separation data;
Tracking the tracking object based on at least one of the first foreground / background separation data or the second foreground / background separation data;
Tracking method. Storing representative color data indicating a representative color for tracking a tracking object;
Generating first foreground / background separation data for separating the foreground data and the background data based on the acquired video data;
Generating second foreground / background separation data based on a region of the color including the representative color in the region of the video data corresponding to the first foreground / background separation data;
Tracking the tracking object based on at least one of the first foreground / background separation data or the second foreground / background separation data;
A tracking program that causes a computer to execute tracking processing.