JP2013164667A - Video retrieval device, method for retrieving video, and video retrieval program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a video retrieval device capable of appropriately and easily retrieving, from a video, a scene in which movement similar to the movement of a subject identified by a user, occurs, and to provide a method for retrieving video, and a video retrieval program.SOLUTION: The video retrieval device includes: a retrieval object area identification unit 18 that identifies position information of an area determined by a user that movement subject to a retrieval occurs and surrounded on a screen by drawing using an input device, as retrieval object area; a movement-occurring section determination unit 19 that determines one section in the video where a series of movement of the retrieval object occurs; and a retrieval processing unit 21 that retrieves, in the retrieval object area identified by the retrieval object area identification unit 18, a video information part where the movement similar to the movement of the subject in the one section determined by the movement-occurring section determination part 19 occurs.

Description

  The present invention relates to a video search apparatus, a video search method, and a video search program for searching a video section (scene) in which a motion similar to the motion of a subject specified by a user has occurred.

  Conventionally, as a technique for searching a scene in which a movement similar to the movement of a designated subject occurs in a video, there are techniques shown in the following Non-Patent Documents 1 and 2, for example.

  Regarding the technique described in Non-Patent Document 1, an image taken in a baseball broadcast will be described as an example.

  First, as a scene to be searched, an image of a short scene (about 70 frames) relating to several movements such as a scene to throw, a scene to swing a bat, and a running scene is prepared.

  Next, the direction (angle) of the motion vector is obtained for the entire screen of the search target video frame, and the frequency of occurrence of each direction is counted. This is repeated for each frame of the image sequence, and a two-dimensional histogram is generated for each video with the direction and time as axes. Then, for each scene related to the motion to be searched, one histogram is selected as a reference, and the corresponding scene is analyzed by shifting in the time axis direction so that the overlap with the histogram is maximized.

  Here, when the histogram is analyzed by a multidimensional scaling method, a scene including the same motion is arranged at close coordinates in the feature space to form a cluster. Since it can be said that scenes belonging to the same cluster contain similar movements, the scenes can be used to search for scenes based on movements.

  Moreover, regarding the technique described in Non-Patent Document 2, an image taken in a soccer broadcast will be described as an example.

  In this technology, a sensor is attached to a soccer player or a ball that is a subject in advance, and the position of the soccer player or the ball is acquired by measuring the sensor position, and is added to the video as time-series metadata. .

  The user who performs the search first selects a desired moving body (player, ball, etc.) from the palette, and draws the locus of the moving body to be searched on the screen on which the video is displayed. At this time, when a plurality of moving objects are selected, the trajectory of each moving object is drawn.

  Then, by matching the drawn trajectory with the time-series data of the moving object, a scene in which a motion similar to the trajectory has occurred can be searched from the search target video.

Zudenek Prohaska, Takashi Ito: "Examination of features for motion recognition in baseball video", IEICE Technical Report. PRMU, Pattern Recognition / Media Understanding, Vol.101, No.125, pp.23-30, 2001 Fumi Yajima, Kazutoshi Kakutani, Katsumi Tanaka, "Football Video Search by Direct Drawing of Motion on Video", IPSJ Research Reports, Database Systems Study Group, Vol.2002, No.41, pp.33-40 , 2002)

  However, the conventional technology described above has the following problems.

  In the technique of Non-Patent Document 1, motion information is obtained from the entire screen and a histogram is created, and a specific method for extracting or specifying motion from the screen has not been studied. For this reason, if the screen contains some movements, or if noises from background movements are included, extraction processing that accurately reflects the movements intended by the user may not be performed. is there.

  In order to use the technique of Non-Patent Document 2, it is necessary to attach a sensor to a person or an object as a subject and acquire time-series position information from this sensor. However, it takes time and effort to mount the sensor. In particular, in the case of a human subject, the presence of the sensor may hinder work and exercise, which is not practical. In addition, when a human limb movement is a search target, a special sensor or a large number of sensors are required, which is not realistic.

  The present invention has been made in view of the above circumstances, and a video search apparatus and video capable of accurately and easily searching a scene in which a motion similar to the motion of a subject specified by a user occurs from a video. It is an object to provide a search method and a video search program.

  In order to solve the above-described problems, the video search apparatus according to the present invention uses the input device on the screen when the user determines that the search target is moving on the screen on which the video is displayed. A search target area designating unit for designating position information of a region surrounded by drawing as a search target area, and a motion generation section determining unit for determining one section in the video in which a series of motions of the search target is generated And a search processing unit for searching for a video information portion in which a movement similar to the movement of the subject in one section determined by the movement generation section determining unit is detected in the search target area specified by the search target area specifying unit; It is characterized by providing.

  In the video search method of the present invention, the video search apparatus is surrounded by a drawing by using the input device on the screen when the user determines that a search target motion is occurring on the screen on which the video is displayed. A search target region specifying step for specifying the position information of the region as a search target region, a motion generation interval determining step for determining one segment in the video where a series of motions of the search target is generated, A search processing step of searching for a video information portion in which a motion similar to the motion of the subject in one section determined in the motion generation section determination step is searched in the search target area specified in the search target area specifying step. It is characterized by.

  According to another aspect of the present invention, there is provided a video search program for causing a computer to execute a function executed by the video search apparatus.

  According to the video search device, the video search method, and the video search program of the present invention, it is possible to accurately and easily search a scene in which a motion similar to the motion of the subject specified by the user occurs from the video. .

It is a block diagram which shows the structure of the video search device by one Embodiment of this invention. It is a flowchart which shows the process which produces | generates motion sequence table information in the image | video search device by one Embodiment of this invention. In the video search device according to one embodiment of the present invention, (a) an example of frame image information in video information to be processed to be processed, and (b) an example of generated motion sequence table information. 5 is a flowchart illustrating an operation when a video search process is executed in a video search device according to an embodiment of the present invention. In a video search device according to an embodiment of the present invention, (a) an example of a screen configuration diagram in which a numerical value of motion angle information is displayed superimposed on frame image information, and (b) a numerical value of motion angle information is converted into color information. FIG. 3 is an example of a screen configuration diagram displayed as a color block superimposed on frame image information. In the video search device according to an embodiment of the present invention, (a) a screen configuration diagram showing a range of interest centered on a position designated on displayed frame image information, and (b) motion angle information corresponding to the range of interest. It is a screen block diagram which shows the state which displayed. 5 is a graph showing the count value of the number of stored motion angle information of each search size area for each frame image information in the video search device according to the embodiment of the present invention. FIG. 6 is a screen display diagram illustrating a state where a search target area input by a user on frame image information is displayed in the video search device according to the embodiment of the present invention. In a video search device according to an embodiment of the present invention, (a) a screen display diagram showing a state in which motion angle information corresponding to a search target area is displayed, and (b) movement of each search size area for each frame image information. It is a graph which shows the count value of the number in which angle information was memorized. 5 is a graph in which motion angle information of a search target area generated by the video search apparatus according to the embodiment of the present invention is classified and counted for each angle range. In the video search apparatus according to the embodiment of the present invention, (a) color conversion information related to one generated frame image information, (b) a histogram image generated by horizontally connecting the color conversion information of each frame image information. Information. FIG. 7 is an explanatory diagram showing processing when a video search process is performed and a similar histogram image information part is extracted from generated histogram image information in the video search device according to the embodiment of the present invention.

<Configuration of Video Retrieval Device According to One Embodiment>
A configuration of a video search apparatus according to an embodiment of the present invention will be described with reference to FIG. The video search device 1 according to the present embodiment includes a video information storage unit 11, a motion detection unit 12, a motion sequence table information storage unit 13, a video display control unit 14, a display unit 15, and a motion visualization information generation unit 16. An input unit 17, a search target area specifying unit 18, a motion generation section determining unit 19, a histogram generating unit 20, a search processing unit 21, and a search result output unit 22.

  The video information storage unit 11 records video information of a tennis match that is video information to be searched. This video information may be a video file recorded in advance, or may be a video stream acquired and recorded from outside as needed.

  The motion detection unit 12 acquires the motion angle information of each position in the frame image information for all frames from the video start time to the video end time constituting the video information recorded in the video information storage unit 11 and stores them. The motion sequence table information having the selected cells is generated.

  The motion sequence table information storage unit 13 stores the motion sequence table information generated by the motion detection unit 12.

  The video display control unit 14 acquires video information stored in the video information storage unit 11 and performs display control.

  The display unit 15 is a screen that displays video information acquired by the video display control unit 14, search results, and the like.

  The motion visualization information generating unit 16 generates motion visualization information for visualizing a place where the motion angle information in the video is generated based on the motion sequence table information stored in the motion sequence table information storage unit 13. Displayed on the display unit 15.

  The input unit 17 is an input device such as a mouse or a touch display operated by the user, and performs an input process for designating an arbitrary position on the display unit 15.

  The search target area designating unit 18 is determined by the user that the movement of the search target is occurring on the screen of the display unit 15 on which the video is displayed, and is surrounded by drawing using the input device of the input unit 17. The position information on the screen of the selected area is designated as the search target area.

  The motion generation section determination unit 19 determines one section in the video where a series of motions to be searched is generated.

  The histogram generation unit 20 changes the motion angle information of each cell corresponding to the search target area specified by the search target area specifying unit 18 in the motion sequence table information stored in the motion sequence table information storage unit 13. The histogram image information shown in the series is generated.

  The search processing unit 21 includes a histogram image information portion corresponding to one section determined by the motion generation section determination unit 19 from histogram image information corresponding to the video start time and end time generated by the histogram generation unit 20. A section having an image information part similar to the above is searched, and is extracted as a section in which a movement similar to the movement to be searched is generated.

  The search result output unit 22 outputs the section extracted by the search of the search processing unit 21 to the display unit 15 as a search result.

<Operation of Video Retrieval Device According to One Embodiment>
As an operation of the video search device 1 according to the present embodiment, video information obtained by shooting a tennis game with a fixed camera is stored in the video information storage unit 11 as a search processing target, and from this video information, a serve scene designated by the user is stored. A video search process that is searched as a search target and displayed on the display unit 15 will be described.

  First, a process for generating motion sequence table information executed as a previous stage of the video search process will be described with reference to the flowchart of FIG.

  In the motion detection unit 12 of the video search device 1, two pieces of continuous frame image information of “video information obtained by shooting a tennis game” which is a search processing target is acquired from the video information storage unit 11 (S 1). A motion vector for each position in the frame image information is detected.

  Specifically, two pieces of continuous frame image information of frame number n and frame number n + 1 are acquired, and image information around the reference coordinate ln in these two pieces of frame image information is compared. Then, the motion vector (optical flow) of the subject with respect to the coordinate ln is detected (S2). In this embodiment, the case where the motion vector of the subject is detected using the block matching method will be described. However, the present invention is not limited to this, and the subject motion vector may be detected using a gradient method, Lucas-Kanade method, or the like.

  If the detected motion vector is lower than a preset threshold, it is ignored as no motion is detected. If the motion vector is equal to or greater than the threshold, the motion angle information of this motion vector is calculated. (S3).

  The motion angle information of the calculated motion vector is stored in motion array table information generated for each frame image information (S4). The motion sequence table information is obtained by dividing two-dimensional sequence table information having cells corresponding to each region in the frame image information divided by a square region whose side is a search size for detecting a motion vector. This is three-dimensional motion arrangement table information having the total number of pieces of frame image information. That is, the motion array table information includes a cell for storing motion vector motion angle information related to positions over all frame image information in the video information.

The total number of frame image information included in the video is fc, the horizontal size of the screen of the display unit 15 is width, the vertical size is height, and the motion vector detection processing is performed on the comparison target region for the reference coordinate ln. When the size is blockSize and the search size is shiftSize, the information amount (total number of cells) of the motion sequence table information is expressed by the following equation (1).

  For example, a motion angle associated with each position for each search size is detected from frame image information 30n as shown in FIG. 3A and frame image information 30n + 1 that follows the frame image information 30n, and a motion angle calculated from each motion vector is detected. FIG. 3B shows a state in which the information is stored in the cell corresponding to the array table information 40n corresponding to the frame image information 30n. In FIG. 3B, the search size is set large to simplify the explanation, and the horizontal size of the screen is divided into 13 pieces according to the search size, and the vertical size is divided into 7 pieces. Show.

  In the example of FIG. 3A, since the act of making a serve is performed on the right side of the screen, a large amount of movement angle information is stored in the right cell also in the corresponding array table information 40n. In addition to the motion of hitting the serve, the motion of other players and the motion of the background are detected, and the calculated motion angle information is stored in the corresponding cell. A value not treated as motion angle information such as null is inserted into a cell corresponding to a position where no motion is detected.

  The processes in steps S2 to S4 described above are repeated for the entire frame image information by repeating the search size shifted from the reference coordinate ln by a search size (loop B).

  Further, loop B processing is executed for all frame image information in the video information (loop A), and motion sequence table information is generated and stored in the motion sequence table information storage unit 13 (S5).

  Next, processing when the video search processing is executed using the stored motion sequence table information will be described with reference to the flowchart of FIG.

  First, video information that is captured in the video information storage unit 11 and that captures a tennis game to be searched is displayed on the display unit 15 under the control of the video display control unit 14. The displayed video is viewed by the user.

  Then, when the user views the video and wants to execute the process with the scene to be served as a search target from the video information, the user uses a playback operation, fast-forward operation, or a seek bar displayed on the screen. The video reproduction control process is executed by the existing technique such as the above-described operation, and the frame image information in any one of the serve scenes in the video information is displayed on the display unit 15.

  Here, when the user views the video, the motion visualization information generating unit 16 generates motion visualization information for visualizing the motion angle information of the motion using the stored motion sequence table information, and the display unit 15 (S11), the user can grasp the movement angle information and can easily find the movement of the subject to be searched.

  For example, as shown in FIG. 5A, the numerical value of the motion angle information in the generated sequence table information 40n information is displayed superimposed on the corresponding location in the frame image information 30n. As shown, motion visualization information 50n obtained by converting the numerical value of the motion angle information of the generated sequence table information 40n into HSV color space information is superimposed and displayed as a color block at a corresponding location in the frame image information 30n. The motion angle information can be visualized.

  Another method for making it easier for the user to find the movement of the subject to be searched will be described with reference to FIGS. For example, as shown in FIG. 6 (a), a fixed range centered on the position indicated by the pointer (position indicated by the arrow) is set as the attention range 60n, and as shown in FIG. 6 (b), the attention range 60n. Only the motion angle information in the motion sequence table information corresponding to is extracted from the motion sequence table information corresponding to each frame image information. For each frame image information, the number of cells in which the movement angle information is stored is counted, and the counted result is displayed in a bar graph in time series as shown in FIG. It is possible to visualize a lot of generated sections in a state in which they are easily visible on the time axis. This bar graph may be displayed on a seek bar displayed on the screen and linearly indicating the start position to the end position of the video. Further, by using the displayed bar graph, the playback position may be skipped to the frame at the time when a large amount of movement has occurred as indicated by the arrow in FIG. 7 by the user's click operation or tap operation.

  Next, in a state where any frame image information in a series of serve scenes that the user wants to search is displayed on the display unit 15, the place where the subject performing the act of hitting the serve is displayed, that is, The area on the screen where it is determined that the movement of the search target has occurred is input by the user as the search target area. The input of the search target area on the screen by the user is executed by being surrounded by a drawn free line by performing an operation such as dragging a mouse as an input device of the input unit 17 or tracing with a finger. The Then, the position information of the input search target area is acquired by the search target area specifying unit 18 to specify the search target area (S12).

  For example, as shown in FIG. 8, the color block displayed by the process of step S11 is referred to, and the search target area 70n is indicated by a free line as shown by a dotted line so as to surround the subject performing the act of serving. It is drawn and input.

  Next, the motion generation section determination unit 19 determines time information of one section in the video information where a series of search target movements occur in the search target area as the motion generation section (S13). For example, in the present embodiment, a section corresponding to a video portion where one serving is performed is determined as a motion generation section. This motion generation section may be determined using a preset constant. For example, the movement generation section is determined as a section corresponding to a total of 2 seconds (corresponding to 60 frames) for 1 second before and after the currently displayed frame. May be. Further, as shown in FIG. 9A, the motion angle information in the motion sequence table information corresponding to the search target area is referred to, and the motion angle information for each frame image information is obtained as shown in FIG. 9B. The stored count value of the number of cells is converted into a bar graph in time series, and the start time corresponding to the frame where a large amount of motion angle information has started to appear before and after the frame image information being displayed is the start point, and the time point corresponding to the reduced frame is displayed. A motion generation section may be determined as the end point. Also, while the video information is being played back, a click operation is performed at the time when the user determines that the movement to be searched has started, and the click operation is released at the time when it is determined that the movement has ended, and the section specified by the user May be determined as the motion generation section.

  Next, the histogram generation unit 20 extracts the cell elements corresponding to the search target area of each frame from the motion sequence table information stored in the motion sequence table information storage unit 13, and displays each element from the video start time to the end time. Histogram image information in which the change information of the appearance frequency of the motion angle information for each angle range at the time is arranged in time series is generated (S14).

  A process for generating histogram image information from motion sequence table information will be described with reference to FIGS. 10 and 11.

  First, regarding the first frame image information, the motion angle information of the motion sequence table information corresponding to the search target area specified in step S12 is classified for each preset angle range and counted.

  For example, for the search target area specified by the frame image information shown in FIG. 8, the motion angle information of the corresponding motion sequence table information acquired as shown in FIG. 9A is obtained for each preset angle range. Appearance frequency is counted. At this time, in order to round off the variation of the motion angle information and combine 0 degrees and 360 degrees, a plurality of angle ranges are set to be overlapped. For example, 24 overlapping angle ranges such as “0 degree to 30 degrees”, “15 degrees to 45 degrees”,... ”345 degrees to 15 degrees” are set, and the frequency of each occurrence is counted. .

  When this counting result is graphed, it is expressed as shown in FIG. In FIG. 10, the horizontal axis indicates the angle range for classifying the motion angle information, and the vertical axis indicates the appearance frequency of each angle range.

  Next, as shown in FIG. 11A, color conversion information in which the counting result shown in FIG. 10 is converted into color shading is generated. In the color conversion information of FIG. 11A, the vertical width corresponds to each angle range for each pixel, the uppermost pixel corresponds to the angular range “0 degree to 30 degrees”, and the lowermost pixel corresponds to the angular range “ The horizontal width is 1 pixel corresponding to the frame, and the higher the frequency of appearance of the angle range, the higher the brightness. Here, since the appearance frequency of the angle range “75 degrees to 90 degrees” is the highest, the pixel corresponding to this angle range is shown in white with the highest luminance.

  In this way, the color conversion information of the search target area is generated for all the frame image information from the start time to the end time of the video information, and these are horizontally connected in time series, and as shown in FIG. Histogram image information 80 having such a width as the number of pixels (fc pixel) corresponding to the frame length of the video is generated. The histogram image information is subjected to normalization processing, and the pixels corresponding to the motion angle information having the highest appearance frequency are expressed so as to have higher luminance.

  Next, in the search processing unit 21, an image information part similar to the histogram image information part corresponding to one section determined by the motion generation section determination unit 19 is selected from the histogram image information 80 generated by the histogram generation unit 20. The section which has is searched (S15).

  In this search process, first, as shown in FIG. 12, the histogram image information portion of the section corresponding to the motion occurrence section determined in step S13 is copied from the histogram image information 80 and acquired as template image information 90. The

Then, matching processing is performed while shifting the template image information 90 with respect to the entire histogram image information 80, and the pixel values of the two image information are compared. The closeness (the degree of image overlap) when comparing images calculated by the matching processing is referred to as similarity here, and is calculated by, for example, an expression for obtaining a correlation as shown by the following expression (2).

  In the above formula (2), T ′ (x ′, y ′) is the pixel value of the template image information 90 at the coordinates (x ′, y ′), and I ′ (x ′, y ′) is the image information to be compared. This is a pixel value (a part of the histogram image information 80 here), and R represents a correlation function.

  For the calculated and normalized similarity, assuming that the similarity is “1” when the image information is completely matched and overlapped, and the similarity when the image information is not matched and does not overlap is “0”, for example, similar The sections determined as having a degree of “0.5” or more are extracted as sections having similar image information portions. A plurality of similar image information portions may be extracted from one histogram image information 80. If multiple sections are extracted at very close positions, they are grouped as one section.

  The video information portion corresponding to the extracted image information portion is predicted to be a video information portion in which a movement similar to the search target movement designated by the user in step S12 occurs.

  Next, in the search result output unit 22, information regarding the video information portion corresponding to the extracted category is output to the display unit 15 as a search result (S16). Display of information related to the video information portion includes display of a thumbnail screen using frame image information in the corresponding video information portion, and corresponding on a seek bar in which each time point from the start to the end in the video information is shown on a straight line. It is performed by displaying the position or reproducing and displaying the corresponding video information portion.

  According to the present embodiment described above, a scene in which a movement similar to the movement of the subject specified by the user occurs in the video can be accurately and easily searched and output based on the user's intention.

  At this time, since the processing is performed based on the search target region input by the user, it is possible to accurately search for the video information portion (scene) based on the movement intended by the user. Thereby, even when a plurality of movements occur in different locations in the same video section or when noise is included, the search process intended by the user can be performed with high accuracy.

  Further, it is not necessary for the user to manually cut out a section corresponding to the motion to be searched for video information in advance, and a scene including an arbitrary motion can be easily searched even from long video information.

  Even when a complicated movement of a subject is a search target, the user can execute the search process only with a simple operation of drawing so as to surround the corresponding subject on the screen. For example, if you want to detect macro movements in which a large number of players move as a whole on the screen, specify the search target area by drawing on the screen so as to encompass and surround the positions of the corresponding large number of players. Search processing is possible only with this. At this time, the histogram generation unit generates the histogram image information based on the appearance frequency of the motion angle information for each angle range of each frame. A search is executed focusing on the trend.

  On the other hand, when searching for a microscopic movement such as the limb of the subject or the movement of the hand, it is only necessary to draw so as to surround only a range corresponding to the movement. At this time, since the template image information used for the matching processing in the search processing unit includes the motion angle information that changes in the motion generation section, the search is executed based on the flow of a series of motions of the search target. The

  In addition, since the location where the motion in the video is generated is visualized based on the motion angle information in the motion visualization information generation unit, even when a motion separate from the motion to be searched for is generated at a close position , The boundaries of individual movements are clearly visualized, making it easier for the user to find the desired movement.

  In addition, by constructing a video search program for causing a computer to execute the functions of the functional units of the video search device of the present embodiment, the computer can also function as a video search device.

DESCRIPTION OF SYMBOLS 1 ... Image | video search device 11 ... Image | video information storage part 12 ... Motion detection part 13 ... Motion sequence table information storage part 14 ... Image | video display control part 15 ... Display part 16 ... Motion visualization information generation part 17 ... Input part 18 ... Search object area | region Designating part 19 ... Generating section determining part 20 ... Histogram generating part 21 ... Search processing part 22 ... Search result output part 30n ... Frame image information 40n ... Motion sequence table information 60n ... Target area 70n ... Search target area 80 ... Histogram image information 90 ... template image information

Claims (7)

On the screen where the video is displayed, the user determines that the movement of the search target is occurring, and uses the input device on the screen to specify the position information of the area surrounded by the drawing as the search target area A search area designating section;
A motion generation section determination unit that determines one section in the video where a series of movements of the search target is generated;
A search processing unit for searching for a video information portion in which a motion similar to the motion of the subject in one section determined by the motion generation section determination unit is detected in the search target region specified by the search target region specifying unit; A video search apparatus characterized by that. The motion angle information at each position for each frame from the start time to the end time of the video is acquired, and block information of different colors is superimposed on the corresponding position on the screen based on the size of the motion angle information. Search the user for display information to be displayed or a graph showing on the time axis the count value of position information that is recognized as the occurrence of movement at each time point from the video start time point to the end time point The video search apparatus according to claim 1, further comprising a motion visualization information generation unit that generates and displays motion visualization information for determining a place where the target motion is occurring. The motion angle information at each position in the specified search target area is acquired for each frame constituting the video, and information indicating the appearance frequency for each angle range of the motion angle information at each position is linked in time series. A histogram generation unit for generating the histogram image information,
The search processing unit is preset with a similarity of a histogram image information portion corresponding to one section determined by the motion generation section determination unit from the histogram image information generated by the motion change information generation unit. The video according to claim 1 or 2, wherein a section having an image information portion determined to be equal to or greater than a threshold is searched and extracted as a section in which a movement similar to the movement of the search target occurs. Search device. Video search device
On the screen where the video is displayed, the user determines that the movement of the search target is occurring, and uses the input device on the screen to specify the position information of the area surrounded by the drawing as the search target area A search area specification step;
A motion generation section determination step for determining one section in the video in which a series of movements of the search target is generated;
A search processing step of searching for a video information portion in which a motion similar to the motion of the subject in one section determined in the motion generation section determination step occurs in the search target area specified in the search target area specifying step. A video search method characterized by the above. The motion angle information at each position for each frame from the start time to the end time of the video is acquired, and block information of different colors is superimposed on the corresponding position on the screen based on the size of the motion angle information. Search the user for display information to be displayed or a graph showing on the time axis the count value of position information that is recognized as the occurrence of movement at each time point from the video start time point to the end time point The video search method according to claim 4, further comprising a motion visualization information generation step of generating and displaying motion visualization information for determining a place where the target motion is occurring. The motion angle information at each position in the specified search target area is acquired for each frame constituting the video, and information indicating the appearance frequency for each angle range of the motion angle information at each position is linked in time series. A histogram generation step for generating the generated histogram image information,
In the search processing step, the similarity of the histogram image information portion corresponding to one section determined in the motion generation section determination step is preset from the histogram image information generated in the motion change information generation step. The video according to claim 4 or 5, wherein a section having an image information portion determined to be equal to or greater than a threshold is searched and extracted as a section in which a movement similar to the movement of the search target occurs. retrieval method.   A video search program for causing a computer to execute a function executed by the video search device according to claim 1.