Single-person sports image summarization method based on key action features and location information

Abstract

The present invention provides a single-person motion image summarization method based on key action features and position information, comprising: obtaining key action frames of a video based on feature extraction and similarity assessment, and obtaining the position information of the person moving therein; performing background modeling and target trajectory extraction on the video; performing collision detection on the position information of the person moving in the key action frames and the target trajectory, directly adding collision-free trajectories to the rearranged trajectory; if there is a collision, performing collision rate detection on the target trajectory and all positions in the key action frames based on a collision rate formula; if the collision rate is less than a threshold, the target trajectory is added to the rearranged trajectory; if the collision rate is greater than or equal to the threshold, the trajectory is discarded; and performing Poisson fusion on the rearranged trajectory and the background image to generate an image summary. The present invention can effectively evaluate motion, and the image summary can also allow users to more intuitively see the shortcomings of their own movements, making it easier to correct them.

Description

Technical Field

[0001] The present invention relates to the technical field of image summarization, and in particular to a single-person motion image summarization method based on key action features and position information. Background Art

[0002] In traditional sports teaching, the evaluation of whether the movement posture is standard depends on the coach's personal experience and vision. There is a lack of quantitative evaluation standards, making it difficult to accurately and effectively evaluate the standardization of the movement and provide reasonable improvement suggestions. Visual-based human posture assessment provides a new interactive method for sports teaching and guidance. It requires an image summarization method combined with human posture recognition to effectively evaluate sports movements in a more professional, intelligent and targeted manner. At the same time, the generated image summary can allow users to more intuitively see the shortcomings of their own movements and facilitate correction of their movements. Summary of the Invention

[0003] Based on the above, in response to the urgent need for intelligent and data-based teaching in schools, this application innovatively combines image summarization technology with human posture recognition technology to design and implement a single-person motion image summarization method based on key action features and position information.

[0004] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0005] In a first aspect, a method for summarizing a single-person motion image based on key action features and position information is provided, comprising:

[0006] Step 1: Extract features from the input single-person motion video V of the person to be detected s to obtain feature information of each frame;

[0007] Step 2: performing a similarity evaluation on the feature information of each frame and the feature information between m standard action key points in a pre-built database, determining m key action frames, and obtaining position information P of the moving person s in the key action frames based on the determined m key action frames; wherein the feature information between the m standard action key points in the database includes distance features and angle features between the m standard action key points;

[0008] Step 3: Perform background modeling and target trajectory extraction on the video V to obtain the background image BG and the trajectory position information of the moving person s;

[0009] Step 4: Rearrange the trajectory according to the spatial relationship between the position information P of the moving person s in the key action frame and the trajectory position information of the moving person s, and move the trajectory position Q that meets the requirements to the desired position. i Add to the position information P to obtain the rearrangement trajectory;

[0010] Step 5: Perform Poisson fusion on the rearranged trajectory and the background image BG to generate an image summary containing all key actions.

[0011] In some embodiments, step 1 includes: inputting a single-person motion video V=[f1, f2, ..., f i ,...,f N ] in the i-th frame f i , f i is the i-th frame w×h image in the video V, w×h is the resolution of the input video, and feature extraction is performed to obtain the feature information of each frame, where the feature information includes the distance feature between M key points [D i，1 , D i，2 ,...,D i，j ,...,D i，M ] and angle features [A i，1 , A i，2 ,...,A i，j ,...,A i，M ]; where D i，j is f i The distance vector between the jth key point and other key points in A i，j is f i The angle vector between the jth key point and other key points in .

[0012] Furthermore, in step 1, feature extraction is performed based on the openpose algorithm, and M is 18.

[0013] In some embodiments, step 2 includes: performing similarity evaluation on the feature information of each frame and the feature information of m standard action key points in the pre-built database, and determining m key action frames [f′1, f′2, ..., f′ j ,...f′ m ], f′ j is an image frame in the video V that has the highest similarity to the j-th key action, j≤m; wherein the feature information between the m standard action key points in the database includes the distance feature and the angle feature between the m standard action key points;

[0014] Based on the determined m key action frames, position information P of the moving person s in the key action frames is obtained = [P1, P2, ..., P j ,...,P m ]; where P j is the moving person s in the image f′ j Location information in .

[0015] In some embodiments, in step 2, the position information P of the moving person s in the key action frame is obtained based on the determined m key action frames. j ,...,P m ],include:

[0016] For m key action frames [f′1, f′2, ..., f′ j ,...f′ m ] Each frame f′ j The moving persons in s use a center point A second-order direction matrix S j and two half-length L j，k Represents a rectangular bounding box of any direction that encloses the moving person s; where k = 0, 1, the second-order direction matrix S j Represents the direction of the two axes of the rectangular bounding box; the second-order direction matrix S j Methods for determining include:

[0017]

[0018]

[0019] where C is the covariance matrix, The key action frame f′ j The x and y components of the coordinates of the contour point of the moving person s are established with the coordinate axis at the upper left corner. express The covariance of E[·] represents the expected calculation, u x ,u y is a vector The mean of

[0020] The eigenvalues and eigenvectors are solved by the covariance matrix C. The direction of the eigenvector corresponding to the maximum eigenvalue is the main axis direction of the bounding box, and the determined second-order direction matrix is obtained.

[0021] Project the pixel coordinates of the moving person s onto the direction vector and find the maximum value v of the x and y components in each direction. k，min and the minimum value v k，max , where k = 0, 1; based on the determined second-order direction matrix and the maximum value v of the x, y components in each direction k，min and the minimum value v k，max , calculate the center point of the bounding box and two half-length L j，k ;

[0022]

[0023]

[0024] Get the position information of the key action frame movement person s P = [P1, P2, ..., P j ,...,P m ]; Each P j It is represented by a rectangular bounding box.

[0025] In some embodiments, step 4 includes:

[0026] The position information P of the moving person s in the key action frame is [P1, P2, ..., P j ,...,P m ]; where P j is the moving person s in the image f′ j Position information in; Position information of moving person s trajectory [Q1, Q2, ..., Q i ,...,Q N ]; where Q i is the image f i Location information of the moving person s;

[0027] According to the position information P of the moving person s in the key action frame and the trajectory position information Q of any moving person s i The spatial relationship between the two is predefined, including:

[0028] a. No collision;

[0029] b. Intersection collision;

[0030] c. Overlap collision;

[0031] Determine the position information P and the trajectory position information Q of the moving person s i Is there a collision?

[0032] a) If the trajectory Q i There is no collision with the position information P, and the trajectory Q i Add to the location information P;

[0033] b) If the trajectory Q of the moving person s i There is a collision with the position information P, calculate the collision rate OR, and determine whether to change the trajectory Q by the collision rate OR and the preset threshold T. i Add to the location information P;

[0034] b1) The trajectory Q with collision rate OR less than threshold T i Add to the location information P;

[0035] b2) If the trajectory Q of the moving person s iIf there is a collision with the position information P, and the collision rate OR is greater than or equal to the threshold T, then this trajectory Q i Discard and do not add to the position information P.

[0036] Furthermore, in some embodiments, the position information P and the trajectory position information Q of the moving person s are determined. i The methods for determining whether there is a collision include:

[0037] For each P in the position information P j and the trajectory Q of the moving person s i , collision detection is performed according to the following formula,

[0038]

[0039] in Indicates existence, a indicates an axis, : indicates that CF(r1, r2) is true, r1, r2 respectively represent the position information P in P j and the trajectory Q of the moving person s i The rectangular bounding box of Respectively represent the projection line segments of r1 and r2 on axis a, express The intersection of represents the empty set;

[0040] If the above conditions are met, it means there is no collision; if the above conditions are not met, it means there is a collision.

[0041] Furthermore, in some embodiments, the collision rate OR is calculated as follows:

[0042]

[0043] Where m is the position information P in P j The number of L j,1 and L j,1 is the rearrangement trajectory P j The half length of the bounding box in both directions, L i,0 and L i,1 is the trajectory Q of the moving person s i The half length of the bounding box in both directions divides the polygonal area where the two bounding boxes intersect into n right triangles, L k,0 and L k,1 are the two right-angled sides of the kth right triangle.

[0044] In a second aspect, the present invention provides a single-person motion image summarization device based on key action features and position information, including a processor and a storage medium;

[0045] The storage medium is used to store instructions;

[0046] The processor is configured to operate according to the instructions to execute the steps of the method according to the first aspect.

[0047] In a third aspect, the present invention provides a storage medium having a computer program stored thereon, which implements the steps of the method described in the first aspect when executed by a processor.

[0048] Beneficial effects: The single-person motion image summarization method based on key action features and position information provided by the present invention has the following advantages: In response to the urgent need for intelligent and data-based school teaching, this application innovatively combines image summarization technology with human posture recognition technology, and designs and implements a single-person motion image summarization method based on key action features and position information. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 The present invention provides a schematic diagram of the overall process of implementing a method for summarizing a single-person motion image based on key action features and position information. DETAILED DESCRIPTION

[0050] The present invention will be further described below in conjunction with the accompanying drawings and examples. The following examples are only used to more clearly illustrate the technical solutions of the present invention and are not intended to limit the scope of protection of the present invention.

[0051] In the description of the present invention, "several" means more than one, "plurality" means more than two, "greater than," "less than," and "exceed" are understood to exclude the number itself, while "above," "below," and "within" are understood to include the number itself. The use of "first" and "second" in the description is solely for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance, implicitly specifying the number of the indicated technical features, or implicitly specifying the order of the indicated technical features.

[0052] In the description of the present invention, reference to terms such as "one embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the exemplary expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0053] Example 1

[0054] A single-person motion image summarization method based on key action features and position information includes:

[0055] Step 1: Extract features from the input single-person motion video V of the person to be detected s to obtain feature information of each frame;

[0056] Step 2: performing a similarity evaluation on the feature information of each frame and the feature information between m standard action key points in a pre-built database, determining m key action frames, and obtaining position information P of the moving person s in the key action frames based on the determined m key action frames; wherein the feature information between the m standard action key points in the database includes distance features and angle features between the m standard action key points;

[0057] Step 3: Perform background modeling and target trajectory extraction on the video V to obtain the background image BG and the trajectory position information of the moving person s;

[0058] Step 4: Rearrange the trajectory according to the spatial relationship between the position information P of the moving person s in the key action frame and the trajectory position information of the moving person s, and move the trajectory position Q that meets the requirements to the desired position. i Add to the position information P to obtain the rearrangement trajectory;

[0059] Step 5: Perform Poisson fusion on the rearranged trajectory and the background image BG to generate an image summary containing all key actions.

[0060] In some embodiments, a single-person motion image summarization method based on key action features and position information, such as Figure 1 As shown, first, the features of the moving people in the input video are extracted and the action similarity is evaluated; then the position information of the moving people in the key action frame is obtained; the background modeling and target trajectory extraction of the input video are performed; the position information of the moving people in the key action frame and the target trajectory are collided with each other, and the trajectory without collision is directly added to the rearranged trajectory. If there is a collision, the collision rate of the target trajectory and all positions in the key action frame is detected according to the collision rate formula. If the collision rate is less than the threshold, the target trajectory is added to the rearranged trajectory. If the collision rate is greater than or equal to the threshold, the trajectory is discarded; finally, the rearranged trajectory is Poisson fused with the background image to generate an image summary containing all key actions. Generate an image summary containing all key actions. The specific implementation includes the following steps:

[0061] Step 1: For a single person motion video V = [f1, f2, ..., f i ,...,f N ] in the i-th frame f i , f i is the i-th frame w×h image in the video V, w×h is the resolution of the input video, and feature extraction is performed to obtain the distance feature and angle feature between key points. Specifically:

[0062] Get fi The key point information of the moving person s, the moving person s includes M key points, and then calculate f i The distance and angle between key points in the image are used to obtain distance features [D i，1 , D i，2 ,...,D i，j ,...,D i，M ] and angle features [A i，1 , A i，2 ,…,A i，j ,…,A i，M ].

[0063] Among them D i，j is f i The distance vector between the jth key point and other key points in .

[0064] Among them A i，j is f i The angle vector between the jth key point and other key points in .

[0065] In some embodiments, feature extraction is performed based on the openpose algorithm. The number of key points in openpose is 18, so M is 18.

[0066] Step 2: Evaluate the similarity between the feature information of each frame and the feature information of m standard action key points in the pre-built database, and obtain m key action frames [f′1, f′2, ..., f′ j ,...f′ m ], f′ j is the frame of image with the highest similarity to the jth key action in the video V, and then obtain the position information P = [P1, P2, ..., P j ,...,P m ], specifically:

[0067] According to the similarity evaluation results of the actions, m key action frames [f′1, f′2, ..., f′ j ,...f′ m ].

[0068] where f′ j It is the frame image with the highest similarity to the j-th key action in the video V.

[0069] Furthermore, the obtaining of the position information P of the moving person s in the key action image specifically includes:

[0070] In order to make the packing of the action person s tighter and make the subsequent collision detection more accurate, the key action frames [f′1, f′2, ..., f′ j,...f′ m ] Each frame f′ j The moving persons in s use a center point A second-order direction matrix S j and two half-length L j，k , where k = 0, 1, represents a rectangular bounding box of any direction that wraps the moving person s, where the second-order direction matrix S j Indicates the directions of the two axes of the bounding box. The specific calculation steps are as follows:

[0071]

[0072]

[0073] where C is the covariance matrix, The key action frame f′ j The x and y components of the coordinates of the pixel point s of the moving person with the upper left corner as the center point to establish the coordinate axis, express The covariance of E[·] represents the expected calculation, u x ,u y is a vector The mean of

[0074] The eigenvalue and eigenvector are solved by the covariance matrix. The direction of the eigenvector corresponding to the maximum eigenvalue is the main axis direction of the bounding box. Assume that the determined direction vector is recorded as Then project the coordinate point onto the direction vector and find the maximum and minimum values of the x and y components in each direction, recorded as v k，min and v k，max , where k = 0, 1. The center of the bounding box can be obtained from formula (3), and the half length of the bounding box can be obtained from formula (4).

[0075]

[0076]

[0077] For key action frames [f′1, f′2, ..., f′ j ,...f′ m ] Each frame f′ j By doing the above processing, the position information P = [P1, P2, ..., P j ,...,P m ].

[0078] Each P j Contains a center point A second-order direction matrix S j and two half-length Lj，k , where k = 0, 1.

[0079] Step 3: Perform background modeling and target trajectory extraction on the input video V to obtain the background image BG and the position information of the trajectory of the moving person s [Q1, Q2, ..., Q i ,...,Q N ], specifically:

[0080] a. Background model initialization: Initialize the background model for each pixel in a single frame image. For each pixel, randomly select the pixel values of its neighboring pixels as its model sample values.

[0081] b. Foreground detection process: Perform foreground object segmentation on subsequent image sequences. The background model stores a sample set for each background point, and then compares each new pixel value with the sample set to determine whether it belongs to the background.

[0082] Furthermore, the background image BG and the position information of the movement person s trajectory [Q1, Q2, ..., Q i ,…,Q N ].

[0083] where Q i is the image f i The location information of the moving person s.

[0084] Step 4: According to the position information P of the moving person s in the key action frame and the trajectory position information of the moving person s [Q1, Q2, ..., Q i ,...,Q N ] to rearrange the trajectory and make the trajectory Q i Add to the position information P, and finally generate the rearrangement trajectory, specifically:

[0085] For position information P and trajectory Q of any moving person s i ,According to their spatial relationship, two trajectories and three relationships are defined, as follows:

[0086] a. No collision.

[0087] b. Intersection collision.

[0088] c. Overlap collision.

[0089] According to the position information P and the trajectory Q of the moving person s i The relationship will meet the required trajectory Q i Add to the location information P, the specific steps are as follows:

[0090] According to formula (5), the position information P and the trajectory Q of the moving person s are determined. iWhether there is a collision, the formula is as follows:

[0091]

[0092] in Indicates existence, a indicates an axis, : indicates that CF(r1, r2) is true, r1, r2 respectively represent the position information P in P j and the trajectory Q of the moving person s i The rectangular bounding box of Respectively represent the projection line segments of r1 and r2 on axis a, express The intersection of Represents the empty set.

[0093] For each P in the position information P j and the trajectory Q of the moving person s i , perform collision detection according to formula (5), if the trajectory Q i There is no collision with the position information P, and the trajectory Q i Added to the location information P.

[0094] Furthermore, if the trajectory Q of the moving person s i If there is a collision with the position information P, it is necessary to use the collision rate formula (6) to determine whether to change the trajectory Q i Add to the position information P, where the collision rate formula is defined as follows:

[0095]

[0096] Where m is the position information P in P j The number of L j，1 and L j，1 is the rearrangement trajectory P j The half length of the bounding box in both directions, L i，0 and L i，1 is the trajectory Q of the moving person s i The half length of the bounding box in both directions divides the polygonal area where the two bounding boxes intersect into n right triangles, L k，0 and L k，1 are the two right-angled sides of the kth right triangle.

[0097] The trajectory Q with collision rate OR less than threshold T i Add to the location information P.

[0098] Furthermore, if the trajectory Q of the moving person s i If there is a collision with the position information P, and the collision rate OR is greater than or equal to the threshold T, then this trajectory Q i Discarded and not added to the rearrangement trajectory P′.

[0099] Step 5: Perform Poisson fusion on the rearranged trajectory and the background image BG to generate an image summary containing all key actions.

[0100] Example 2

[0101] In a second aspect, this embodiment provides a single-person motion image summarization device based on key action features and position information, including a processor and a storage medium;

[0102] The storage medium is used to store instructions;

[0103] The processor is configured to operate according to the instructions to execute the steps of the method according to embodiment 1.

[0104] Example 3

[0105] In a third aspect, this embodiment provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described in Example 1 are implemented.

[0106] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0107] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0108] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1The function specified in one or more boxes.

[0109] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0110] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A single-person motion image summarization method based on key action features and position information, characterized by: The steps include: Step 1: Extract features from the input single-person motion video V of the person to be detected s to obtain feature information of each frame; Step 2: performing a similarity evaluation on the feature information of each frame and the feature information between m standard action key points in a pre-built database, determining m key action frames, and obtaining position information P of the moving person s in the key action frames based on the determined m key action frames; wherein the feature information between the m standard action key points in the database includes distance features and angle features between the m standard action key points; Step 3: Perform background modeling and target trajectory extraction on the video V to obtain the background image BG and the trajectory position information of the moving person s; Step 4: Rearrange the trajectory according to the spatial relationship between the position information P of the moving person s in the key action frame and the trajectory position information of the moving person s, and add the trajectory position that meets the requirements to the position information P to obtain the rearranged trajectory, including: judging the position information P and the trajectory position information Q of the moving person s i Is there a collision? a) If Q i There is no collision with the position information P, and Q i Add to the location information P; b) If the Q of the athlete s i There is a collision with the position information P, calculate the collision rate OR, and determine whether to move Q based on the collision rate OR and the preset threshold T. i Add to the location information P; b1) The collision rate OR is less than the Q threshold T i Add to the location information P; b2) If the collision rate OR is greater than or equal to the threshold T, then Q i Discard and do not add to the position information P; Step 5: Perform Poisson fusion on the rearranged trajectory and the background image BG to generate an image summary containing all key actions.

2. The single-person motion image summarization method based on key action features and position information according to claim 1, characterized in that: Step 1 includes: inputting a single person motion video V = [f1, f2, ..., f i ,…,f N ] in the i-th frame f i , f i is the i-th frame w×h image in the video V, w×h is the resolution of the input video, and feature extraction is performed to obtain the feature information of each frame, where the feature information includes the distance feature between M key points [D i,1 ,D i,2 ,…,D i,j ,…,D i,M ] and angle features [A i,1 ,A i,2 ,…,A i,j ,…,A i,M ]; where D i,j is f i The distance vector between the jth key point and other key points in A i,j is f i The angle vector between the jth key point and other key points in .

3. The single-person motion image summarization method based on key action features and position information according to claim 2, characterized in that: In step 1, feature extraction is performed based on the openpose algorithm, and M is 18.

4. The single-person motion image summarization method based on key action features and position information according to claim 2, characterized in that: Step 2 includes: performing similarity evaluation on the feature information of each frame and the feature information of m standard action key points in the pre-built database, and determining m key action frames [f′1, f′2, …, f′ j ,…f′ m ], f′ j is an image frame in the video V that has the highest similarity to the j-th key action, j≤m; wherein the feature information between the m standard action key points in the database includes the distance feature and the angle feature between the m standard action key points; Based on the determined m key action frames, position information P of the moving person s in the key action frames is obtained = [P1, P2, ..., P j ,…,P m ]; where P j is the moving person s in the image f′ j Location information in .

5. The single-person motion image summarization method based on key action features and position information according to claim 4 is characterized in that: In the step 2, the position information P of the moving person s in the key action frame is obtained based on the determined m key action frames. j ,…,P m ],include: For m key action frames [f′1,f′2,…,f′ j ,…f′ m ] Each frame f′ j The moving persons in s use a center point A second-order direction matrix S j and two half-length L j,k Represents a rectangular bounding box of any direction that encloses the moving person s; where k = 0, 1, the second-order direction matrix S j Represents the direction of the two axes of the rectangular bounding box; the second-order direction matrix S j Methods for determining include: where C is the covariance matrix, The key action frame f′ j The upper left corner is the center point to establish the coordinate axis of the moving person s contour point x, y components, express The covariance of E[·] represents the expected calculation, u x ,u y is a vector The mean of The eigenvalues and eigenvectors are solved by the covariance matrix C. The direction of the eigenvector corresponding to the maximum eigenvalue is the main axis direction of the bounding box, and the determined second-order direction matrix is obtained. Project the pixel coordinates of the moving person s onto the direction vector and find the maximum value v of the x and y components in each direction. k,min and the minimum value v k,max , where k = 0, 1; based on the determined second-order direction matrix and the maximum value v of the x, y components in each direction k,min and the minimum value v k,max , calculate the center point of the bounding box and two half-length L j,k ; Get the position information of the key action frame motion person s P=[P1,P2,…,P j ,…,P m ]; Each P j It is represented by a rectangular bounding box.

6. The single-person motion image summarization method based on key action features and position information according to claim 5, characterized in that: The step 4 comprises: The position information of the moving person s in the key action frame is P = [P1, P2, ..., P j ,…,P m ]; where P j is the moving person s in the image f′ j Position information in; Position information of moving person s trajectory [Q1,Q2,…,Q i ,…,Q N ]; where Q i is the image f i The trajectory position information of the moving person s; According to the position information P of the moving person s in the key action frame and the trajectory position information Q of any moving person s i The spatial relationship between the two is predefined, including: a. No collision; b. Intersection collision; c. Overlap collision.

7. The single-person motion image summarization method based on key action features and position information according to claim 6, characterized in that: Determine the position information P and the trajectory position information Q of the moving person s i The methods for determining whether there is a collision include: For each P in the position information P j Q&A with sports staff i , collision detection is performed according to the following formula, in Indicates existence, a indicates an axis, : indicates that CF(r1,r2) is true, r1,r2 respectively represent the position information P in P j Q and sports personnel i The rectangular bounding box of Respectively represent the projection line segments of r1 and r2 on axis a, express The intersection of represents the empty set; If the above conditions are met, it means there is no collision; if the above conditions are not met, it means there is a collision.

8. The single-person motion image summarization method based on key action features and position information according to claim 6, characterized in that: The calculation method of collision rate OR is: Where m is the position information P in P j The number of L j,1 and L j,1 is the rearrangement trajectory P j The half length of the bounding box in both directions, L i,0 and L i,1 It is the Q of sports personnel i The half length of the bounding box in both directions divides the polygonal area where the two bounding boxes intersect into n right triangles, L k,0 and L k,1 are the two right-angled sides of the kth right triangle.

9. A single-person motion image summarization device based on key action features and position information, characterized in that: including processor and storage medium; The storage medium is used to store instructions; The processor is configured to operate according to the instructions to execute the steps of the method according to any one of claims 1 to 8.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images