Action matching method, device, equipment and storage medium

Abstract

The present application provides an action matching method, device, equipment and storage medium, which relates to the field of action recognition technology. The method includes: obtaining an image to be matched with an action to be matched; using a target skeleton model corresponding to a preset standard action to identify the skeleton points of the image to be matched, and obtaining the target skeleton point position; obtaining an action template of the preset standard action, wherein the action template of the preset standard action stores: the standard skeleton point position of the preset standard action; according to the target skeleton point position and the standard skeleton point position, calculating the target matching score of the action to be matched and the preset standard action, the target matching score is used to characterize the matching degree between the action to be matched and the preset standard action. The present application can meet the training needs of action requirements, reduce human participation, improve the quality of training assessment, and ensure the consistency of assessment standards.

Description

Technical Field

[0001] The present invention relates to the field of motion recognition technology, and in particular to a motion matching method, device, equipment and storage medium. Background Art

[0002] With the rapid development of science and technology, people's requirements for life are getting higher and higher, and more and more industries are in greater and greater demand for professionals with professional skills, which makes the training and assessment of professionals particularly important.

[0003] The training and assessment of professionals, especially those with action requirements, is mostly conducted through teacher lectures. Moreover, when professionals enter the work, they also need to undergo regular training and assessment of professional skills. Moreover, the training and assessment are also conducted manually by experienced personnel.

[0004] At present, the training and assessment of professionals who require action still mostly rely on human experience. Due to the limited number of teachers, it is unable to meet the existing training needs. Secondly, the experience and skill levels of teachers vary, making it difficult to ensure the quality of their teaching. At the same time, the skill assessment of professionals is still based on human judgment by examiners, and the consistency of their evaluation standards cannot be ensured. Summary of the invention

[0005] The purpose of the present invention is to provide a motion matching method, device, equipment and storage medium to address the deficiencies in the above-mentioned prior art, so as to solve the problem that the existing solutions cannot meet the motion training needs and the training quality is difficult to guarantee.

[0006] To achieve the above purpose, the technical solution adopted by the embodiment of the present invention is as follows:

[0007] In a first aspect, an embodiment of the present invention provides an action matching method, comprising:

[0008] Acquire an image to be matched with an action to be matched;

[0009] Using a target skeleton model corresponding to a preset standard action, the skeleton points of the image to be matched are identified to obtain the position of the target skeleton points;

[0010] Acquire an action template of the preset standard action, wherein the action template of the preset standard action stores: a standard skeleton point position of the preset standard action;

[0011] According to the target skeleton point position and the standard skeleton point position, a target matching score between the action to be matched and the preset standard action is calculated, and the target matching score is used to characterize the matching degree between the action to be matched and the preset standard action.

[0012] Optionally, calculating the target matching score of the action to be matched and the preset standard action according to the target skeleton point position and the standard skeleton point position includes:

[0013] According to the positions of the multiple groups of first skeleton points in the target skeleton point positions and the positions of the multiple groups of first skeleton points in the standard skeleton point positions, the matching scores of the multiple groups of first skeleton points are calculated respectively; wherein the first skeleton points are: skeleton points with skeleton linkage in the human body;

[0014] Calculating a first matching score according to the matching scores of the multiple groups of first skeleton points;

[0015] According to the positions of the plurality of groups of second skeletal points in the target skeletal point positions and the positions of the plurality of groups of second skeletal points in the standard skeletal point positions, the matching scores of the plurality of groups of second skeletal points are calculated respectively; wherein the second skeletal points are skeletal points in the human body without skeletal linkage;

[0016] Calculating a second matching score according to the matching scores of the plurality of groups of second skeleton points;

[0017] According to the positions of the plurality of groups of third skeleton points in the target skeleton point positions and the positions of the plurality of groups of third skeleton points in the standard skeleton point positions, the matching scores of the plurality of groups of third skeleton points are calculated respectively; the third skeleton points are: skeleton points with non-fixed distances;

[0018] Calculating a third matching score according to the matching scores of the plurality of groups of third skeleton points;

[0019] The target matching score is calculated according to the first matching score, the second matching score and the third matching score.

[0020] Optionally, the calculating the matching scores of the multiple groups of first skeleton points respectively according to the positions of the multiple groups of first skeleton points in the target skeleton point positions and the positions of the multiple groups of first skeleton points in the standard skeleton point positions includes:

[0021] According to the first recognition vector formed by the position of each group of first skeleton points in the target skeleton point position and the first standard vector formed by the position of each group of first skeleton points in the standard skeleton point position, the matching score of each group of first skeleton points is calculated.

[0022] Optionally, the step of calculating the matching score of each group of first skeleton points according to the first recognition vector formed by the position of each group of first skeleton points in the target skeleton point position and the first standard vector formed by the position of each group of first skeleton points in the standard skeleton point position comprises:

[0023] An inner product operation is performed on the first recognition vector and the first standard vector to obtain a matching score for each group of first skeleton points.

[0024] Optionally, the calculating the matching scores of the multiple groups of second skeleton points respectively according to the positions of the multiple groups of second skeleton points in the target skeleton point positions and the positions of the multiple groups of second skeleton points in the standard skeleton point positions includes:

[0025] The matching score of each group of second skeleton points is calculated based on the second recognition vector formed by the position of each group of second skeleton points in the target skeleton point position and the second standard vector formed by the position of each group of second skeleton points in the standard skeleton point position.

[0026] Optionally, the step of calculating the matching score of each group of second skeleton points according to the second recognition vector formed by the position of each group of second skeleton points in the target skeleton point position and the second standard vector formed by the position of each group of second skeleton points in the standard skeleton point position comprises:

[0027] An inner product operation is performed on the second recognition vector and the second standard vector to obtain a matching score for each group of second skeleton points.

[0028] Optionally, the calculating the matching scores of the multiple groups of third skeleton points respectively according to the positions of the multiple groups of third skeleton points in the target skeleton point positions and the positions of the multiple groups of third skeleton points in the standard skeleton point positions includes:

[0029] The matching score of each group of third bone points is calculated based on the first bone distance between the positions of each group of third bone points in the target bone point position and the second bone distance between the positions of each group of third bone points in the standard bone point position.

[0030] Optionally, calculating the matching score of each group of third skeletal points according to a first skeletal distance between the positions of each group of third skeletal points in the target skeletal point positions and a second skeletal distance between the positions of each group of third skeletal points in the standard skeletal point positions comprises:

[0031] Calculating a first relative distance of each group of third bone points according to the first bone distance and a first unit distance between the positions of the reference bones in the target bone point positions;

[0032] Calculate the second relative distance of each group of third skeleton points according to the second skeleton distance and the second unit distance between the positions of the reference skeletons in the standard skeleton point positions;

[0033] The matching score of each group of third skeleton points is calculated according to the first relative distance of each group of third skeleton points and the second relative distance of each group of third skeleton points.

[0034] Optionally, calculating the matching score of each group of third skeleton points according to the first relative distance of each group of third skeleton points and the second relative distance of each group of third skeleton points includes:

[0035] Calculating the first relative distance and the second relative distance, and calculating the bone distance deviation of each group of third bone points;

[0036] The matching score of each group of third bone points is calculated according to the ratio of the bone distance deviation and the second relative distance.

[0037] Optionally, calculating the first matching score according to the matching scores of the multiple groups of first skeleton points includes:

[0038] Determining a proportionality coefficient of each group of first skeleton points according to the matching score of each group of first skeleton points;

[0039] A weighted sum operation is performed according to the matching scores of the multiple groups of first skeleton points and the proportional coefficient of each group of first skeleton points to obtain the first matching score.

[0040] Optionally, calculating the second matching score according to the matching scores of the plurality of groups of second skeleton points includes:

[0041] Determining a proportional coefficient of each group of second skeleton points according to the matching score of each group of second skeleton points;

[0042] A weighted sum operation is performed according to the matching scores of the multiple groups of second skeleton points and the proportional coefficient of each group of second skeleton points to obtain the second matching score.

[0043] Optionally, calculating a third matching score according to the matching scores of the plurality of groups of third skeleton points includes:

[0044] Determining a proportional coefficient of each group of third skeletal points according to the matching score of each group of third skeletal points;

[0045] A weighted sum operation is performed according to the matching scores of the multiple groups of third skeleton points and the proportional coefficient of each group of third skeleton points to obtain the third matching score.

[0046] Optionally, the method further includes:

[0047] Determining a plurality of first weight coefficients of the plurality of groups of first skeleton points respectively according to the matching scores of the plurality of groups of first skeleton points;

[0048] Determining a plurality of second weight coefficients of the plurality of groups of second skeleton points respectively according to the matching scores of the plurality of groups of second skeleton points;

[0049] Determine a plurality of third weight coefficients of the plurality of groups of third skeletal points respectively according to third identification vectors of the plurality of groups of third skeletal points in the target skeletal points and third standard vectors of the plurality of groups of third skeletal points in the standard skeletal points;

[0050] The calculating the target matching score according to the first matching score, the second matching score and the third matching score includes:

[0051] Calculating a total matching score according to the first matching score, the second matching score and the third matching score;

[0052] Calculating a total weight coefficient according to the plurality of first weight coefficients, the plurality of second weight coefficients, and the plurality of third weight coefficients;

[0053] The target matching score is calculated according to the total matching score and the total weight coefficient.

[0054] Optionally, calculating the target matching score according to the total matching score and the total weight coefficient includes:

[0055] The target matching score is calculated according to the ratio of the total matching score to the total weight coefficient.

[0056] Optionally, the method further includes:

[0057] Determining whether the target matching score is greater than or equal to a preset benchmark matching score;

[0058] If the target matching score is greater than or equal to the reference matching score, it is determined that the action to be matched has been completed and the preset action requirement corresponding to the action to be matched is met.

[0059] Optionally, the step of using a skeleton model corresponding to a preset standard action to identify a target skeleton point position in the image to be matched includes:

[0060] Using the target skeleton model, identifying skeleton points of multiple frames of the to-be-matched images having the to-be-matched action, and obtaining multiple groups of first skeleton point positions, each group of first skeleton point positions including: positions of skeleton points identified from one frame of the to-be-matched image;

[0061] The plurality of groups of first skeleton point positions are averaged to obtain the target skeleton point position.

[0062] Optionally, before obtaining the action template of the preset standard action, the method further includes:

[0063] Acquire a standard image having the preset standard action;

[0064] Using the target skeleton model, identifying the skeleton points of the standard image to obtain the positions of the standard skeleton points;

[0065] The standard skeleton point position is stored in the action template of the preset standard action.

[0066] Optionally, the adopting the target skeleton model to identify skeleton points of the standard image to obtain the positions of the standard skeleton points includes:

[0067] Using the target skeleton model, identifying skeleton points of multiple frames of the standard image having the preset standard action, and obtaining multiple groups of second skeleton point positions, each group of second skeleton point positions includes: skeleton point positions identified from one frame of the standard image;

[0068] The plurality of groups of second skeleton point positions are averaged to obtain the standard skeleton point position.

[0069] In a second aspect, an embodiment of the present application further provides an action matching device, including:

[0070] A first acquisition module, used for acquiring a to-be-matched image having an to-be-matched action;

[0071] A recognition module is used to identify the skeleton points of the image to be matched by using the target skeleton model corresponding to the preset standard action to obtain the position of the target skeleton point;

[0072] A second acquisition module is used to acquire an action template of the preset standard action, wherein the action template of the preset standard action stores: a standard skeleton point position of the preset standard action;

[0073] A calculation module is used to calculate the target matching score of the action to be matched and the preset standard action according to the target skeleton point position and the standard skeleton point position, wherein the target matching score is used to characterize the matching degree between the action to be matched and the preset standard action.

[0074] In a third aspect, an embodiment of the present application further provides a computer device, comprising: a memory and a processor, wherein the memory stores a computer program executable by the processor, and when the processor executes the computer program, it implements any action matching method provided in the first aspect above.

[0075] In a fourth aspect, an embodiment of the present application further provides a storage medium having a computer program stored thereon, and when the computer program is read and executed, any action matching method provided in the first aspect above is implemented.

[0076] The beneficial effects of this application are:

[0077] In the action matching method, device, equipment and storage medium provided by the present application, a target skeleton model corresponding to a preset standard action can be used to identify the skeleton points of the image to be matched with the action to be matched, obtain the target skeleton point position, and obtain the action template of the preset standard action, the action template of the preset standard action stores: the standard skeleton point position of the preset standard action, and then according to the target skeleton point position and the standard skeleton point position, calculate the target matching score of the action to be matched and the preset standard action, the target matching score is used to characterize the matching degree of the action to be matched with the preset standard action. By adopting the action matching method, the matching score of the action to be matched can be obtained by comparing the target skeleton point position identified from the image to be matched with the standard skeleton point position in the action template of the preset standard action, without the need for manual training and assessment, the training and assessment of personnel with action requirements can be realized, the experience of manual experience can be effectively reduced, and it is not limited by the number of teachers, and can meet the increasing training needs, and the reduction of manual participation can ensure the teaching quality as much as possible, and also reduce manual judgment, effectively ensuring the consistency of assessment standards. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0079] Figure 1 A schematic diagram of a scenario of a training and assessment system to which the action matching method provided in an embodiment of the present application is applied;

[0080] Figure 2 A flowchart of an action matching method provided in an embodiment of the present application;

[0081] Figure 3 A schematic diagram of a torso skeleton model used in an embodiment of the present application;

[0082] Figure 4 A schematic diagram of a process for identifying the position of a target skeleton point in an action matching method provided in an embodiment of the present application;

[0083] Figure 5 A schematic diagram of a process for generating an action template for a standard action in an action matching method provided in an embodiment of the present application;

[0084] Figure 6A schematic diagram of a process for identifying the position of a standard skeleton point in an action matching method provided in an embodiment of the present application;

[0085] Figure 7 A schematic diagram of a process for calculating a target matching score in an action matching method provided in an embodiment of the present application;

[0086] Figure 8 A schematic diagram of a flow chart of calculating a matching score of a third skeleton point in an action matching method provided in an embodiment of the present application;

[0087] Fig. 9 A schematic diagram of another process of calculating a target matching score in an action matching method provided in an embodiment of the present application;

[0088] Fig.10 A schematic diagram of a flow chart of an action specification recognition in an action matching method provided in an embodiment of the present application;

[0089] Fig.11 A schematic diagram of an action matching device provided in an embodiment of the present application;

[0090] Fig.12 A schematic diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0091] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments.

[0092] The motion matching method provided in the embodiment of the present application can be applied to a training and assessment system to perform motion training and / or motion assessment on professionals with motion requirements. The professionals with motion requirements can be, for example, flight attendants, dancers, gymnasts, and other personnel with motion requirements. Figure 1 A schematic diagram of a scenario of a training and assessment system to which the action matching method provided in the embodiment of the present application is applied. Figure 1 As shown, the training and assessment system may include: a computer device 01 and a camera 02. The camera 02 may be a main camera or a network camera. The computer device 01 may be connected to the camera 02 by wired communication or by wireless communication. Figure 1 The connection lines in the figure are only for illustration of communication connection and do not limit the specific communication method.

[0093] It should be noted that camera 02 can be a computer camera, a computer eye, an electronic eye, etc. Its specific product form can be an infrared camera or a depth camera. It can be a camera with an image acquisition sensor, or an image acquisition device such as a terminal device with a camera or a camera, etc. The embodiment of the present application does not limit the specific type of the camera.

[0094] An action matching application is installed on the computer device 01. By running the action matching application, the action matching methods provided in the following embodiments of the present application are executed, and the corresponding action matching results are output.

[0095] The action matching method provided in the present application is illustrated below through multiple examples with reference to the accompanying drawings.

[0096] Figure 2 A flowchart of an action matching method provided in an embodiment of the present application is provided. The action matching method can be implemented by the computer device 01 in the above-mentioned training and assessment system through software and / or hardware. Figure 2 As shown, the method may include:

[0097] S201: Acquire a to-be-matched image having an to-be-matched action.

[0098] In a possible implementation, the image to be matched can be obtained from a preset storage location during the action recognition process, or the image to be matched captured by the camera can be directly obtained from the camera. The preset storage location can be, for example, a local storage location on a computer, or a cloud storage location such as a cloud server.

[0099] The image to be matched may be an image captured by the camera when the trainee performs the action to be matched. That is, the computer device may obtain the offline image to be matched captured by the camera, or may obtain the offline image to be matched from the preset storage location when the camera stores the image to be matched in the preset storage location after capturing the image to be matched.

[0100] S202: using a target skeleton model corresponding to a preset standard action, identifying skeleton points of the image to be matched, and obtaining the position of the target skeleton points.

[0101] When the computer runs the action matching application, it can call a preset human posture recognition program, use the skeleton target skeleton model, identify the skeleton points of the image to be matched, and obtain the target skeleton point position. The preset human posture recognition program can be an openpose recognition program or a kinect recognition program.

[0102] Before identifying the skeletal points, the skeletal model with the movable joints of the preset standard action can be determined as the target skeletal model according to the preset standard action. For example, the skeletal model with the movable joints of the preset standard action can be determined as the target skeletal model from a preset skeletal model library. The preset standard action can be a standard action corresponding to the action to be matched, and the movable joint can be a skeletal joint that needs to be movable to complete the preset standard action. The preset skeletal model library may include: multiple groups of skeletal models, such as a trunk skeletal model, i.e., a skeletal model including trunk joint points, a hand skeletal model, i.e., a skeletal model including hand joint points, and a foot skeletal model, i.e., a skeletal model including foot joint points.

[0103] Therefore, before determining the target skeleton model, the active joint of the preset standard action can be determined first, and then the target skeleton model can be obtained based on the active joint. For example, if the active joint is a trunk joint, the target skeleton model can be a trunk skeleton model; if the active joint is a hand joint, the target skeleton model can be a hand skeleton model; if the active joint is a foot joint, the target skeleton model can be a foot skeleton model.

[0104] The trunk skeleton model may be a 25-point model, that is, the trunk skeleton model includes 25 skeleton points. Figure 3 Schematic diagram of the trunk skeleton model used in the embodiment of the present application. Figure 3 As shown, the torso skeleton model may include 25 skeleton points from skeleton point 0 to skeleton point 24. Among them, skeleton point 0 is a nose skeleton point, skeleton point 1 is a spine shoulder skeleton point, skeleton point 2 is a right shoulder skeleton point, skeleton point 3 is a right elbow skeleton point, skeleton point 4 is a right wrist skeleton point, skeleton point 5 is a left shoulder skeleton point, skeleton point 6 is a left elbow skeleton point, skeleton point 7 is a left wrist skeleton point, skeleton point 8 is a spine base skeleton point, skeleton point 9 is a right waist skeleton point, skeleton point 10 is a right knee skeleton point, skeleton point 11 is a right ankle skeleton point, and skeleton point 12 is a left waist skeleton point. , bone point 13 is the left knee bone point, bone point 14 is the left ankle bone point, bone point 15 is the right eye bone point, bone point 16 is the left eye bone point, bone point 17 is the right ear bone point, bone point 18 is the left ear bone point, bone point 19 is the left sole bone point, bone point 20 is the left toe bone point, bone point 21 is the left heel bone point, bone point 22 is the right sole bone point, bone point 23 is the right toe bone point, and bone point 24 is the right heel bone point.

[0105] It should be pointed out that the torso skeleton model can also be a skeleton model with other forms of bone point annotations. Figure 3 This is just an example of a torso skeleton model.

[0106] The hand skeleton model may be a model including key points of the hand skeleton, which may be a 22-point model, that is, the hand skeleton model includes 22 skeleton points. This application does not provide a specific description of the diagram of the 22-point model.

[0107] After determining the target skeleton model, the target skeleton model can be used to identify the skeleton points of the image to be matched, and the position of each skeleton point in the target skeleton model in the image to be matched is obtained, which is the target skeleton point position. It should be noted that, for the convenience of unified description, the skeleton points involved in the embodiments of the present application can also be described as skeleton key points or key skeleton points and other similar descriptions, and the present application does not specifically limit this.

[0108] S203, obtaining an action template of the preset standard action, wherein the action template of the preset standard action stores: a standard skeleton point position of the preset standard action.

[0109] For example, the action target of the preset standard action can be obtained from the preset action module panel based on the identification of the preset standard action, such as the name or serial number. The action template library can store an action template of at least one standard action, and the action template of each standard action can store the standard skeleton point position of each standard action.

[0110] The standard bone point position of the preset standard action may be the standard position of each bone point in the target bone model under the preset standard action.

[0111] S204. Calculate a target matching score between the action to be matched and the preset standard action according to the target skeleton point position and the standard skeleton point position, wherein the target matching score is used to represent a matching degree between the action to be matched and the preset standard action.

[0112] For example, the data difference between the target bone point position and the standard bone point position can be calculated, and the target matching score can be calculated based on the data difference. The larger the data difference between the target bone point position and the standard bone point, the closer the action to be matched is to the preset standard action, and the larger the target matching score is, which can represent the higher the matching degree between the action to be matched and the preset standard action. The smaller the data difference between the target bone point position and the standard bone point, the farther the action to be matched is from the preset standard action, and the smaller the target matching score is, which can represent the lower matching degree between the action to be matched and the preset standard action.

[0113] After obtaining the target matching score, it can be stored and / or output for display as the action score of the user corresponding to the action to be matched, or a judgment can be made based on the target matching score and a preset benchmark matching score to obtain the action assessment result of the user corresponding to the action to be matched.

[0114] The action matching method provided in the embodiment of the present application can use the target skeleton model corresponding to the preset standard action to identify the skeleton points of the image to be matched with the action to be matched, obtain the target skeleton point position, and obtain the action template of the preset standard action, the action template of the preset standard action stores: the standard skeleton point position of the preset standard action, and then calculates the target matching score of the action to be matched and the preset standard action according to the target skeleton point position and the standard skeleton point position, and the target matching score is used to characterize the matching degree of the action to be matched with the preset standard action. Using this action matching method, the matching score of the action to be matched can be obtained by comparing the target skeleton point position identified from the image to be matched with the standard skeleton point position in the action template of the preset standard action, without the need for manual training and assessment, the training and assessment of personnel with action requirements can be realized, the experience of manual experience can be effectively reduced, and it is not limited by the number of teachers, and can meet the increasing training needs, and the reduction of manual participation can ensure the teaching quality as much as possible, and also reduce manual judgment, effectively ensuring the consistency of assessment standards.

[0115] Moreover, in the action matching method, action templates of standard actions are adopted, which avoids the use of action matching models based on neural networks for matching, thereby avoiding the high cost of model training, reducing the resource cost required for action matching, avoiding the rigid defects of model matching, and improving the applicability of action matching.

[0116] Optionally, in other possible implementations, the embodiment of the present application may also provide an implementation example of identifying the position of a target skeleton point in an action matching method. Figure 4 A schematic diagram of a process for identifying the position of a target skeleton point in an action matching method provided in an embodiment of the present application, such as Figure 4 As shown, in the method shown above, in S202, a target skeleton model corresponding to a preset standard action is used to identify skeleton points of the image to be matched, and the position of the target skeleton point is obtained, which may include:

[0117] S401, using the target skeleton model, identifying skeleton points of multiple frames of the to-be-matched images having the to-be-matched action, and obtaining multiple groups of first skeleton point positions.

[0118] Each group of first skeleton point positions includes: positions of skeleton points identified from a frame of the image to be matched.

[0119] In a possible implementation, the image to be matched may be acquired at a preset frame rate, which may be an image acquisition frame rate of a camera, for example, 20 frames per second. In this way, multiple frames of the image to be matched may be acquired.

[0120] After acquiring multiple frames of the image to be matched, the target skeleton model can be used to identify the skeleton points of each frame of the image to be matched, and a group of first skeleton point positions can be obtained. By identifying the skeleton points of multiple frames of the image to be matched, multiple groups of first skeleton point positions can be obtained. Each group of first skeleton point positions may include: the positions of the skeleton points identified from one frame of the image to be matched.

[0121] Assuming that the target skeleton model is a 25-point torso skeleton model, the set of first skeleton point positions may include: positions of 25 skeleton points identified from a frame of image to be matched.

[0122] S402: Averaging the multiple groups of first skeleton point positions to obtain the target skeleton point position.

[0123] When the skeleton points are identified by using multiple frames of images to be matched and the multiple groups of first skeleton point positions are obtained, the multiple groups of first skeleton point positions can be smoothed. For example, the target skeleton point position can be obtained by using the average position line of the skeleton points according to the multiple groups of first skeleton point positions. The average position line is to average the positions of the same skeleton points in the multiple groups of first skeleton point positions to obtain the target position P(x′ n ,y′ n ), in this way, the target position of multiple skeleton points can be obtained, which is the target skeleton point position. For example, the positions of the nasal skeleton points identified in multiple frames of images to be matched can be averaged to obtain the target position of the nasal skeleton point.

[0124] For example, the target position of the skeleton point, that is, the position of the skeleton point after smoothing, can be calculated using the following formula (1).

[0125]

[0126] Among them, x i is the coordinate position of the bone point identified from the i-th frame to be matched image, y i is another coordinate position of the skeleton point identified from the i-th frame of the image to be matched. n is the total number of frames of the image to be matched that participate in the skeleton point identification.

[0127] The action matching method provided in this embodiment can obtain the target bone point position by averaging multiple groups of first bone point positions identified from multiple frames of images to be matched, thereby avoiding inaccurate identified bone point positions caused by image frame skipping, improving the recognition accuracy of the target bone point position, and accurately evaluating the matching action and the preset standard action, thereby improving the reliability and accuracy of the evaluation of action training and assessment.

[0128] Optionally, in other possible implementations, the embodiment of the present application may also provide an implementation example of an action template for a standard action in an action matching method. Figure 5 A schematic diagram of a flow chart of generating an action template of a standard action in an action matching method provided in an embodiment of the present application, such as Figure 5 As shown, before obtaining the action template of the preset standard action in S203 in the method shown above, the method may further include:

[0129] S501: Acquire a standard image having the preset standard action.

[0130] For example, the standard image can be obtained during the process of saving the action template of the preset standard action. The specific implementation process of obtaining the standard image can be similar to the implementation process of obtaining the image to be matched, and the specific description is referred to above, which will not be repeated here.

[0131] The difference is that the action in the image to be matched is the action to be matched, and whether the action is standard is unknown, and it is necessary to execute the action matching method provided in the embodiment of the present application for identification, while the action in the standard image is a standard action, which can be an image captured by a person with rich action experience such as an action training teacher, or a humanoid robot pre-configured with standard actions performing standard actions.

[0132] S502: using the target skeleton model, identifying skeleton points of the standard image to obtain the positions of the standard skeleton points.

[0133] After acquiring the standard image, the target skeleton model can be used to identify the skeleton points of the standard image to obtain the position of each skeleton point in the target skeleton model in the standard image, which is the standard skeleton point position.

[0134] S503: storing the standard skeleton point position in the action template of the preset standard action.

[0135] When the standard skeleton point positions are obtained through identification in S502, they can be stored in the action template of the preset standard action.

[0136] When standard images of multiple standard actions are obtained, the target skeleton model corresponding to each standard action can be used to identify the skeleton points of the standard images of each standard action to obtain the standard skeleton point positions of each standard action. In this way, the standard skeleton point positions of multiple standard actions can be achieved.

[0137] The standard skeleton point position of a standard action can be stored in the action template of the standard action, so for multiple standard actions, multiple action templates can be obtained.

[0138] After obtaining multiple action templates, multiple action templates storing the standard skeleton point positions corresponding to the standard actions can also be stored in a preset standard action template library, so as to facilitate obtaining the action templates for action matching during the action recognition process. The standard template library can be located in a local storage device of a computer device executing the action matching method, or in a remote storage device such as a cloud storage device that communicates with the computer device.

[0139] It should be pointed out that the implementation Figure 5 The computer device for saving the action template of the standard action can be the same device as the computer device for executing the action matching method, or can be a different computer device, and this application does not limit this.

[0140] Before performing the action matching method, the standard bone point positions are obtained by performing bone point recognition on the standard image, and then stored in the action template of the preset standard action, thereby realizing the acquisition of the action template. It is convenient to compare the standard bone point positions in the action template with the target bone point positions obtained in the action matching process, and then realize the matching of the action to be matched and the preset standard action, thereby realizing action training and assessment, reducing the resource cost required for action matching, and improving the flexibility of action matching. If a newly added standard action needs to be matched, it is only necessary to pre-recognize the standard image of the newly added standard action to obtain the action template of the newly added standard action without re-training the model.

[0141] For example, the embodiments of the present application may also provide a possible implementation method for identifying the positions of standard skeleton points in an action matching method. Figure 6 A flow chart of identifying the position of a standard skeleton point in an action matching method provided in an embodiment of the present application, as shown in / 6, uses the target skeleton model in S502 of the method shown above to identify the skeleton points of the standard image to obtain the position of the standard skeleton point, which may include:

[0142] S601, using the target skeleton model, identifying skeleton points of multiple frames of the standard image having the preset standard action, and obtaining multiple groups of second skeleton point positions.

[0143] Each group of second skeleton point positions includes: skeleton point positions identified based on a frame of the standard image.

[0144] After acquiring multiple frames of standard images, the target skeleton model can be used to identify the skeleton points of each frame of standard images to obtain a set of second skeleton point positions. By identifying the skeleton points of multiple frames of action images, multiple sets of second skeleton point positions can be obtained. Each set of second skeleton point positions may include: the positions of the skeleton points identified from a frame of standard images.

[0145] S602: Averaging the multiple groups of second skeleton point positions to obtain the standard skeleton point position.

[0146] When the skeleton points are identified by using multiple frames of standard images to obtain the multiple sets of second skeleton point positions, the multiple sets of second skeleton point positions can be smoothed. For example, the standard skeleton point positions can be obtained by using the average position line of the skeleton points according to the multiple sets of second skeleton point positions. The average position line is obtained by averaging the positions of the same skeleton points in the multiple sets of second skeleton point positions to obtain the standard positions of the skeleton points. In this way, the standard positions of multiple skeleton points can be obtained, which are the standard skeleton point positions.

[0147] For example, the positions of the nasal bone points identified in 10 consecutive frames of standard images may be averaged to obtain the center of the nasal bone points in the 10 consecutive frames of standard images, which is the standard position of the nasal bone points.

[0148] In the action matching method provided in this embodiment, the standard bone point position can be obtained by averaging multiple groups of second bone point positions identified from multiple frames of standard images, which can avoid inaccurate identification of standard bone point positions caused by image frame skipping, improve the recognition accuracy of standard bone point positions, accurately match and judge the matching action and the preset standard action, and improve the reliability and accuracy of action training and assessment.

[0149] Optionally, the embodiment of the present application may further provide an action matching method, which is illustrated below with reference to multiple implementation methods for calculating the matching score. Figure 7 A schematic diagram of a process for calculating a target matching score in an action matching method provided in an embodiment of the present application. Figure 7 As shown, optionally, as shown in S204 above, calculating the target matching score of the action to be matched and the preset standard action according to the target skeleton point position and the standard skeleton point position may include:

[0150] S701, respectively calculating matching scores of the multiple groups of first skeleton points according to the positions of the multiple groups of first skeleton points in the target skeleton point position and the positions of the multiple groups of first skeleton points in the standard skeleton point position.

[0151] The multiple groups of first skeleton points may be multiple groups of skeleton points within the human body, that is, skeleton points with skeleton linkage within the human body.

[0152] Specifically, the matching scores of a group of first skeleton points can be calculated based on the positions of a group of first skeleton points in the target skeleton point position and the positions of the group of first skeleton points in the standard skeleton point position. In this way, for multiple groups of first skeleton points, the matching scores of the multiple groups of first skeleton points can be obtained.

[0153] The target skeleton model is as above Figure 3 Taking the trunk skeleton model shown in FIG. 1 as an example, the matching scores of 20 groups of first skeleton points can be calculated by using S701. The 20 groups of first skeleton points can be, for example, 20 groups of skeleton points from sequence number 1 to sequence number 20 as shown in Table 1 below.

[0154] Table 1

[0155] Serial number Bone Point 1 Bone Point 2 1 8 1 2 1 0 3 1 2 4 2 3 5 3 4 6 1 5 7 5 6 8 6 7 9 8 9 10 9 10 11 10 11 12 8 12 13 12 13 14 13 14 15 11 22 16 11 23 17 11 24 18 14 19 19 14 20 20 14 21

[0156] In this way, by executing S701, matching scores of 20 groups of first skeleton points are obtained.

[0157] In a possible implementation example, in S701, respectively calculating the matching scores of the multiple groups of first skeleton points according to the positions of the multiple groups of first skeleton points in the target skeleton point position and the positions of the multiple groups of first skeleton points in the standard skeleton point position may include:

[0158] The matching score of each group of first skeleton points is calculated based on the first recognition vector formed by the position of each group of first skeleton points in the target skeleton point position and the first standard vector formed by the position of each group of first skeleton points in the standard skeleton point position.

[0159] For example, the relative angle of each group of first skeleton points may be calculated according to the first recognition vector and the first standard vector.

[0160] Optionally, an inner product operation is performed on the first recognition vector and the first standard vector to obtain a matching score for each group of first skeleton points.

[0161] If the first recognition vector is a and the first standard vector is b, the matching score of each group of first skeleton points can be calculated using the following formula (2). The matching score of the i-th first skeleton point in the multiple groups of first skeleton points can be expressed as S′ i1 , for example, can be calculated using the following formula (2).

[0162] S′ i1 =(a, b) i =cos<a,b> i ...Formula (2)

[0163] S702: Calculate a first matching score according to the matching scores of the multiple groups of first skeleton points.

[0164] The following are examples of possible implementations:

[0165] In a possible implementation, the matching scores of the multiple groups of first skeleton points may be accumulated to obtain the first matching score.

[0166] In another possible implementation, the first matching score may be obtained by performing a weighted sum operation on the matching scores of the multiple groups of first skeleton points according to the scale coefficients of the multiple groups of first skeleton points.

[0167] In this implementation, each group of first skeleton points may be pre-configured with a corresponding proportional coefficient. The matching score of each group of first skeleton points is weighted according to the proportional coefficient of each group of first skeleton points, and then the weighted matching scores of multiple groups of first skeleton points are summed to obtain the first matching score.

[0168] In another possible implementation, the proportional coefficient of each group of first bone points can be determined based on the matching score of each group of first bone points, and then a weighted sum operation can be performed based on the matching scores of the multiple groups of first bone points and the proportional coefficient of each group of first bone points to obtain the first matching score.

[0169] For example, the matching score of each group of first skeleton points can be compared with a preset first score threshold, and the proportional coefficient of each group of first skeleton points can be determined according to the comparison result. For example, the proportional coefficient of each group of first skeleton points can be determined by the following formula (3) and formula (4), and then the matching score of each group of first skeleton points is weighted according to the proportional coefficient of each group of first skeleton points to obtain the target matching score of each group of first skeleton points.

[0170] S i1 =S′ i1 *k 11 , satisfying S′ i1 <0.707......Formula (3)

[0171] S i1 =S′ i1 *k 12 , satisfying S′ i1 ≥0.707......Formula (4)

[0172] Among them, S i1 is the target matching score of the first skeleton point in the i-th group, k 11 k is a proportional coefficient determined when the matching score is less than the first score threshold, which can be, for example, 1, k 12 When the matching score is greater than or equal to the first score threshold, a proportionality coefficient is determined, which may be 1.2, for example.

[0173] When the target matching score of each group of first skeleton points is obtained by the above method, the first matching score Score1 can also be calculated by using formula (5).

[0174]

[0175] S703: Calculate matching scores of the multiple groups of second skeleton points according to the positions of the multiple groups of second skeleton points in the target skeleton point position and the positions of the multiple groups of second skeleton points in the standard skeleton point position.

[0176] The plurality of groups of second skeleton points may be a plurality of groups of skeleton points inside the human body, that is, skeleton points without skeleton linkage inside the human body.

[0177] Specifically, the matching scores of a group of second skeleton points can be calculated based on the positions of a group of second skeleton points in the target skeleton point position and the positions of the group of second skeleton points in the standard skeleton point position. In this way, for multiple groups of second skeleton points, the matching scores of the multiple groups of second skeleton points can be obtained.

[0178] The target skeleton model is as above Figure 3 Taking the trunk skeleton model shown in FIG. 1 as an example, the matching scores of 22 groups of second skeleton points can be calculated by using S703. The 22 groups of second skeleton points can be, for example, 22 groups of skeleton points from sequence number 1 to sequence number 22 as shown in Table 2 below.

[0179] Table 2

[0180]

[0181]

[0182] In this way, by executing S703, matching scores of 22 groups of second skeleton points are obtained.

[0183] In a possible implementation example, in S703, respectively calculating the matching scores of the multiple groups of second skeleton points according to the positions of the multiple groups of second skeleton points in the target skeleton point position and the positions of the multiple groups of second skeleton points in the standard skeleton point position may include:

[0184] The matching score of each group of second skeleton points is calculated based on the second recognition vector formed by the position of each group of second skeleton points in the target skeleton point position and the second standard vector formed by the position of each group of second skeleton points in the standard skeleton point position.

[0185] For example, the relative angle of each group of second skeleton points may be calculated according to the second recognition vector and the second standard vector.

[0186] Optionally, an inner product operation is performed on the second recognition vector and the second standard vector to obtain a matching score for each group of second skeleton points.

[0187] S704: Calculate a second matching score according to the matching scores of the multiple groups of second skeleton points.

[0188] The matching score of the i-th second skeleton point in the plurality of groups of second skeleton points can be expressed as S′i2 .

[0189] The following are examples of possible implementations:

[0190] In a possible implementation, the matching scores of the multiple groups of second skeleton points may be accumulated to obtain the second matching score.

[0191] In another possible implementation, the matching scores of the multiple groups of second skeleton points may be weighted and calculated according to the scale coefficients of the multiple groups of second skeleton points to obtain the second matching score.

[0192] In this implementation, each group of second skeleton points may be pre-configured with a corresponding proportional coefficient. The matching score of each group of second skeleton points is weighted according to the proportional coefficient of each group of second skeleton points, and then the weighted matching scores of multiple groups of second skeleton points are summed to obtain the second matching score.

[0193] In another possible implementation, the proportional coefficient of each group of second skeleton points can be determined based on the matching score of each group of second skeleton points, and then a weighted sum operation can be performed based on the matching scores of the multiple groups of second skeleton points and the proportional coefficient of each group of second skeleton points to obtain the second matching score.

[0194] For example, the matching score of each group of second skeletal points can be compared with a preset second score threshold, and the proportional coefficient of each group of second skeletal points can be determined according to the comparison result. For example, the proportional coefficient of each group of second skeletal points can be determined using the following formulas (6) and (7), and then the matching score of each group of second skeletal points is weighted according to the proportional coefficient of each group of second skeletal points to obtain the target matching score of each group of second skeletal points.

[0195] S i2 =S′ i2 *k 21 , satisfying S′ i2 <0.707......Formula (6)

[0196] S i2 =S′ i2 *k 22 , satisfying S′ i2 ≥0.707......Formula (7)

[0197] Among them, S i2 is the target matching score of the second skeleton point in the i-th group, k 21 k is a proportional coefficient determined when the matching score is less than the second score threshold, which can be, for example, 1, k 22 When the matching score is greater than or equal to the second score threshold, a proportionality coefficient is determined, which may be 1.2, for example.

[0198] When the target matching score of each group of second skeleton points is obtained by the above method, the second matching score Score2 can also be calculated by formula (8).

[0199]

[0200] S705. Calculate matching scores of the multiple groups of third skeleton points according to the positions of the multiple groups of third skeleton points in the target skeleton point position and the positions of the multiple groups of third skeleton points in the standard skeleton point position.

[0201] The plurality of groups of third skeleton points may be: a plurality of groups of skeleton points with non-fixed distances, that is, skeleton points with distances between which change as the action changes.

[0202] Specifically, the target distance of a group of third skeleton points can be obtained according to the position of a group of third skeleton points in the target skeleton point position, and the standard distance of the group of third skeleton points can be obtained according to the position of the group of third skeleton points in the standard skeleton point position. According to the target distance and the standard distance, the matching score of the group of third skeleton points is obtained by calculating the distance deviation. In this way, for multiple groups of third skeleton points, the matching scores of the multiple groups of third skeleton points can be obtained.

[0203] Continue with the target bone model as above Figure 3 Taking the trunk skeleton model shown in the figure as an example, the matching scores of 5 groups of third skeleton points can be calculated by using S705. The 5 groups of second skeleton points can be, for example, 5 groups of skeleton points from sequence number 1 to sequence number 5 as shown in Table 3 below.

[0204] Table 3

[0205] Serial number Bone Point 1 Bone Point 2 1 11 14 2 4 9 3 7 12 4 1 8 5 0 1

[0206] In a possible implementation example, the above S705, according to the positions of the multiple groups of third skeleton points in the target skeleton point position and the positions of the multiple groups of third skeleton points in the standard skeleton point position, respectively calculates the matching scores of the multiple groups of third skeleton points, which may include:

[0207] The matching score of each group of third bone points is calculated based on the first bone distance between the positions of each group of third bone points in the target bone point position and the second bone distance between the positions of each group of third bone points in the standard bone point position.

[0208] During the implementation process, the direct distance deviation of the first bone distance and the second bone distance can be calculated, or the corresponding relative distances can be calculated respectively according to the first bone distance and the second bone distance, and the relative distance deviation can be calculated according to the corresponding relative distances, and then the matching score of each group of third bone points can be calculated according to the relative distance deviation.

[0209] S706: Calculate a third matching score according to the matching scores of the multiple groups of third skeleton points.

[0210] The matching score of the i-th third skeleton point in the multiple groups of third skeleton points can be expressed as S′ i3 .

[0211] The following are examples of possible implementations:

[0212] In a possible implementation, the matching scores of the multiple groups of third skeleton points may be accumulated to obtain the third matching score.

[0213] In another possible implementation, the matching scores of the multiple groups of third skeleton points may be weighted and calculated according to the scale coefficients of the multiple groups of third skeleton points to obtain the third matching score.

[0214] In this implementation, each group of third skeletal points may be pre-configured with a corresponding proportional coefficient. The matching score of each group of third skeletal points is weighted according to the proportional coefficient of each group of third skeletal points, and then the weighted matching scores of multiple groups of third skeletal points are summed to obtain the third matching score.

[0215] In another possible implementation, the proportional coefficient of each group of third skeletal points can be determined based on the matching score of each group of third skeletal points, and then a weighted sum operation can be performed based on the matching scores of the multiple groups of third skeletal points and the proportional coefficient of each group of third skeletal points to obtain the third matching score.

[0216] For example, the inner product of the third identification vector a′ of each group of third skeletal points in the target skeletal point position and the third standard vector b′ of each group of third skeletal points in the standard skeletal point position can be compared with a preset inner product threshold, and the proportional coefficient of each group of third skeletal points can be determined according to the comparison result. For example, the proportional coefficient of each group of third skeletal points can be determined by the following formulas (9) and (10), and then the matching score of each group of third skeletal points can be weighted according to the proportional coefficient of each group of third skeletal points to obtain the target matching score of each group of third skeletal points.

[0217] S i3 =S′ i3 *k 31 , satisfying (a′, b′) i <-0.25||(a′, b′)i >0.25......Formula (9)

[0218] S i3 =S′ i3 *k 32 , satisfying (a′, b′) i >-0.25&(a′,b′) i <0.25......Formula (10)

[0219] Among them, S i3 is the target matching score of the third skeleton point in the i-th group, k 31 k is a proportionality coefficient determined when the inner product of the third recognition vector and the third standard vector is less than a negative preset inner product threshold value such as -0.25, or greater than a preset inner product threshold value such as 0.25, which can be 3, for example. 32 When the inner product of the third recognition vector and the third standard vector is greater than a negative preset inner product threshold and less than the preset inner product threshold, the proportional coefficient is determined, which may be 4, for example.

[0220] When the target matching score of each group of third skeleton points is obtained by the above method, the third matching score Score3 can also be calculated by formula (11).

[0221]

[0222] S707: Calculate the target matching score according to the first matching score, the second matching score, and the third matching score.

[0223] For example, in one possible implementation, the first matching score, the second matching score, and the third matching score may be accumulated to obtain a total matching score, which is the target matching score; in another possible implementation, the first matching score, the second matching score, and the third matching score may be accumulated to obtain a total matching score, and the target matching score may be calculated based on the total matching score and the preset coefficients corresponding to the respective matching scores.

[0224] In the action matching method provided in the embodiment of the present application, the first matching score, the second matching score, and the third matching score can be calculated respectively from the three aspects of the internal skeleton of the human body, the external skeleton of the human body, and the bone distance according to the target skeleton point position and the standard skeleton point position, and then the target matching score is calculated. That is, in addition to using the vectors of the internal skeleton points of the human body, the vectors and bone distances of the external skeleton points of the human body are added to effectively improve the accuracy of action recognition, and several groups of bone point distances are added to distinguish some small action changes, thereby improving the accuracy and precision of action matching. Therefore, the action matching method can make the matching dimensions of the target matching score more, and more accurately reflect the matching degree between the action to be matched and the preset standard action, so as to facilitate accurate action assessment during training and assessment.

[0225] Moreover, in the above-mentioned action matching method, the matching scores of different first skeletal points, the matching scores of the second skeletal points, and the matching scores of the third skeletal points are multiplied by different proportional coefficients, respectively, so that the differences in details that are concerned in the assessment can be revealed through the scores, thereby expanding the differences in subtle movements.

[0226] Optionally, in some other possible examples, the embodiments of the present application may also provide an action matching method, which is explained as follows by calculating the matching score of the third skeleton point. Figure 8 The following is a flow chart of calculating the matching score of the third skeleton point in an action matching method provided in an embodiment of the present application. Figure 8 As shown, optionally, as shown in S705 above, respectively calculating the matching scores of the multiple groups of third skeleton points according to the positions of the multiple groups of third skeleton points in the target skeleton point position and the positions of the multiple groups of third skeleton points in the standard skeleton point position may include:

[0227] S801. Calculate the first relative distance of each group of third bone points according to the first bone distance and the first unit distance between the positions of the reference bones in the target bone point position.

[0228] S802: Calculate the second relative distance of each group of third bone points according to the second bone distance and the second unit distance between the positions of the reference bones in the standard bone point positions.

[0229] For example, the reference bone can be the left and right shoulder bone points, and the first unit distance can be the distance between the left and right shoulder bone points in the target bone point position, that is, the shoulder width of the user who performs the action to be matched in the image to be matched; the second unit distance can be the distance between the left and right shoulder bone points in the target bone point position, that is, the shoulder width of the user who performs the preset standard action in the standard image.

[0230] Before describing, the calculation method of the distance between two bone points can be illustrated by an example, for example, it can be calculated using the following formula (12).

[0231]

[0232] Among them, (x1, y1) represents the position of the A skeleton point, and (x1, y1) represents the position of the B skeleton point.

[0233] For example, the first relative distance may be calculated according to the ratio of the first bone distance to the first unit distance, and the second relative distance may be calculated according to the ratio of the second bone distance to the second unit distance.

[0234] S803. Calculate the matching score of each group of third skeleton points according to the first relative distance of each group of third skeleton points and the second relative distance of each group of third skeleton points.

[0235] In a possible implementation, the deviation of the relative distance can be calculated based on the first relative distance of each group of third skeletal points and the second relative distance of each group of third skeletal points, and then the matching score of each group of third skeletal points can be calculated based on the deviation of the relative distance.

[0236] For example, as shown in S803 above, the matching score of each group of third skeleton points is calculated according to the first relative distance of each group of third skeleton points and the second relative distance of each group of third skeleton points, including:

[0237] Calculating the first relative distance and the second relative distance, and calculating the bone distance deviation of each group of third bone points;

[0238] The matching score of each group of third bone points is calculated based on the ratio of the bone distance deviation and the second relative distance.

[0239] The bone distance deviation may be, for example, a difference between the first relative distance and the second relative distance, that is, a difference in relative distances.

[0240] For example, when the first bone distance, the first reference distance, the second bone distance, and the second reference distance are known, formula (13) can be used to calculate and obtain the matching score S of each group of third bone points: i3 .

[0241]

[0242] Wherein, distance_b is the first bone distance, that is, the distance of the third bone point in the target bone point identified in the image to be matched, distance_b stdis the first reference distance, i.e., the distance between the left and right shoulder bone points in the target bone points identified in the image to be matched. distance_a is the second bone distance, i.e., the distance between the third bone points in the action template. std The second reference distance is the distance between the left and right shoulder bone points in the action template.

[0243] is the first relative distance, is the second relative distance, is the above bone distance deviation.

[0244] By using the relative distance method, the relative distance of the skeleton points in the image and the relative distance of the skeleton points in the standard action are calculated respectively, and then the matching score of each group of the third skeleton points is calculated, which can ensure the calculation accuracy of the third matching score and more accurately reflect the distance matching degree between the action to be matched and the preset standard action, which is convenient for accurate action assessment during training and assessment. In addition, the use of the concept of relativity improves the applicability of action matching recognition on the basis of improving the accuracy rate of action recognition.

[0245] On the basis of any of the above embodiments, the embodiment of the present application may also provide another implementation of the action matching method, which is described below through the specific process of calculating the target matching score. Fig. 9 This is a schematic diagram of another process for calculating a target matching score in an action matching method provided in an embodiment of the present application. Fig. 9 As shown, the method may also include:

[0246] S901. Determine a plurality of first weight coefficients of the plurality of groups of first skeleton points respectively according to the matching scores of the plurality of groups of first skeleton points.

[0247] For example, based on the comparison result of the matching score of each group of first skeleton points with the first score threshold, the weight coefficient corresponding to the comparison result can be determined as the weight parameter of each group of first skeleton points, thereby obtaining the multiple first weight coefficients.

[0248] For example, the multiple first weight coefficients may be determined using formulas (14) and (15).

[0249] n i1 =1, satisfying S′ i1 <0.707......Formula (14)

[0250] n i1 =3, satisfying S′ i1 ≥0.707......Formula (15)

[0251] Among them, 0.707 is an example of the first score threshold, and it can also be other values. The values ​​of the weight coefficients corresponding to different comparison results are also only possible examples, and this application does not limit this.

[0252] S902: Determine a plurality of second weight coefficients of the plurality of groups of second skeleton points respectively according to the matching scores of the plurality of groups of second skeleton points.

[0253] Using the above 901, multiple second weight coefficients can be obtained, and the second weight coefficient can be expressed as n i2 .

[0254] S903, determining a plurality of third weight coefficients of the plurality of groups of third skeletal points respectively according to the third identification vectors of the plurality of groups of third skeletal points in the target skeletal points and the third standard vectors of the plurality of groups of third skeletal points in the standard skeletal points.

[0255] For example, a vector inner product can be performed based on the third recognition vector and the third standard vector, and the vector inner product can be compared with a preset inner product threshold to determine the weight coefficient corresponding to the comparison result as the weight parameter of each group of third bone points, thereby obtaining the multiple third weight coefficients.

[0256] For example, the multiple second weight coefficients may be determined using formulas (16) and (17).

[0257] m i =10, satisfying (a′, b′) i <-0.25||(a′, b′) i >0.25......Formula (16)

[0258] m i =12, satisfying (a′, b′) i >-0.25&(a′,b′) i <0.25......Formula (17)

[0259] Among them, 0.25 is an example of the inner product threshold, and it can also be other values. The values ​​of the weight coefficients corresponding to different comparison results are also only possible examples, and this application does not limit this.

[0260] Accordingly, calculating the target matching score according to the first matching score, the second matching score and the third matching score in S707 may include:

[0261] S904: Calculate a total matching score according to the first matching score, the second matching score, and the third matching score.

[0262] S905: Calculate a total weight coefficient according to the multiple first weight coefficients, the multiple second weight coefficients, and the multiple third weight coefficients.

[0263] S906: Calculate the target matching score according to the total matching score and the total weight coefficient.

[0264] For example, the target matching score can be calculated based on the ratio of the total matching score to the total weight coefficient. Of course, other methods can also be used for calculation, such as combining other parameters for calculation, and the ratio is only one possible calculation method.

[0265] When multiple first weight coefficients, multiple second weight coefficients, and multiple third weight coefficients are obtained, the following formulas (18)-(20) can be used to calculate the sum of the first weight coefficients, the sum of the second weight coefficients, and the sum of the third weight coefficients, respectively, and then the total weight coefficient is obtained.

[0266]

[0267]

[0268]

[0269] According to the total matching score and the total weight coefficient, formula (21) is used to summarize and obtain the target matching score.

[0270]

[0271] In the action matching method provided in this embodiment, the weight coefficient of each matching score can make the difference in details that are concerned in the assessment appear through the score, thereby expanding the difference in subtle actions. The target matching score is calculated by the ratio of the total matching score and the total weight coefficient, and the change range of the target matching score is constrained between 0 and 100%, that is, the score is normalized, which is convenient for better matching of actions and improves the applicability of action matching.

[0272] On the basis of any of the above embodiments, the embodiment of the present application may also provide another implementation of the action matching method, which is described below by the specific process of the operation after calculating the target matching score. Fig.10 The following is a flow chart of an action specification recognition process in an action matching method provided in an embodiment of the present application. Fig.10 As shown, the method may also include:

[0273] S1001. Determine whether the target matching score is greater than or equal to a preset benchmark matching score.

[0274] S1002: If the target matching score is greater than or equal to the benchmark matching score, it is determined that the action to be matched has been completed and the preset action requirement corresponding to the action to be matched is met.

[0275] The preset benchmark matching score may be, for example, 85%. When the target matching score reaches the benchmark matching score, that is, greater than or equal to the benchmark matching score, it can be determined that the action to be matched has been completed and meets the preset action requirements, and is a standard action.

[0276] If the target matching score is lower than the benchmark matching score, it can be determined that the action to be matched is not completed, or the completed action does not meet the preset action requirements.

[0277] If it is determined through judgment that the action is not completed, the action of obtaining the image to be matched can be re-executed, and then the action matching method can be re-executed until it is detected that the action to be matched is completed.

[0278] During the actual training or assessment process, after judging the target matching score of the action to be matched and the preset benchmark matching score, the matching result of the action to be matched can also be output. The matching result may include, for example: the target matching score, whether the action to be matched is completed, and whether the preset action requirements are met and other indication information.

[0279] By comparing the target matching score with the preset benchmark matching score, a corresponding comparison judgment result is obtained and output, so that the user can timely and accurately know the action result of the action to be matched.

[0280] The following describes the apparatus, device, storage medium, etc. used to execute the action matching method provided by the present application. The specific implementation process and technical effects are described above and will not be repeated below.

[0281] Fig.11 A schematic diagram of an action matching device provided in an embodiment of the present application is shown in FIG. Figure 7 As shown, the action matching device 1100 may include:

[0282] The first acquisition module 1101 is used to acquire a to-be-matched image having an to-be-matched action.

[0283] The recognition module 1102 is used to use the target skeleton model corresponding to the preset standard action to recognize the skeleton points of the image to be matched, and obtain the position of the target skeleton point.

[0284] The second acquisition module 1103 is used to acquire the action template of the preset standard action, and the action template of the preset standard action stores: the standard skeleton point position of the preset standard action.

[0285] The calculation module 1104 is used to calculate the target matching score of the action to be matched and the preset standard action according to the target skeleton point position and the standard skeleton point position, and the target matching score is used to represent the matching degree between the action to be matched and the preset standard action.

[0286] Optionally, the calculation module 1104 is specifically used to calculate the matching scores of the multiple groups of first skeletal points according to the positions of the multiple groups of first skeletal points in the target skeletal point positions and the positions of the multiple groups of first skeletal points in the standard skeletal point positions; wherein the first skeletal points are: skeletal points with skeletal linkage in the human body; the first matching scores are calculated according to the matching scores of the multiple groups of first skeletal points; the matching scores of the multiple groups of second skeletal points are calculated according to the positions of the multiple groups of second skeletal points in the target skeletal point positions and the positions of the multiple groups of second skeletal points in the standard skeletal point positions; Among them, the second bone point is: a bone point in the human body where there is no bone linkage; the second matching score is calculated based on the matching scores of the multiple groups of second bone points; the matching scores of the multiple groups of third bone points are calculated respectively based on the positions of the multiple groups of third bone points in the target bone point position and the positions of the multiple groups of third bone points in the standard bone point position; the third bone point is: a bone point with an unfixed distance; the third matching score is calculated based on the matching scores of the multiple groups of third bone points; the target matching score is calculated based on the first matching score, the second matching score and the third matching score.

[0287] Optionally, the calculation module 1104 is specifically used to calculate the matching score of each group of first bone points based on the first recognition vector formed by the position of each group of first bone points in the target bone point position and the first standard vector formed by the position of each group of first bone points in the standard bone point position.

[0288] Optionally, the calculation module 1104 is specifically configured to perform an inner product operation on the first recognition vector and the first standard vector to obtain a matching score for each group of first skeleton points.

[0289] Optionally, the calculation module 1104 is specifically used to calculate the matching score of each group of second bone points based on a second identification vector formed by the position of each group of second bone points in the target bone point position and a second standard vector formed by the position of each group of second bone points in the standard bone point position.

[0290] Optionally, the calculation module 1104 is specifically configured to perform an inner product operation on the second recognition vector and the second standard vector to obtain a matching score for each group of second skeleton points.

[0291] Optionally, the calculation module 1104 is specifically used to calculate the matching score of each group of third bone points based on the first bone distance between the positions of each group of third bone points in the target bone point position and the second bone distance between the positions of each group of third bone points in the standard bone point position.

[0292] Optionally, the calculation module 1104 is specifically used to calculate the first relative distance of each group of third bone points based on the first bone distance and the first unit distance between the position of the reference bone in the target bone point position; calculate the second relative distance of each group of third bone points based on the second bone distance and the second unit distance between the position of the reference bone in the standard bone point position; calculate the matching score of each group of third bone points based on the first relative distance of each group of third bone points and the second relative distance of each group of third bone points.

[0293] Optionally, the calculation module 1104 is specifically used to calculate the bone distance deviation of each group of third bone points based on the first relative distance and the second relative distance; and calculate the matching score of each group of third bone points based on the bone distance deviation and the ratio of the second relative distance.

[0294] Optionally, the calculation module 1104 is specifically used to determine the proportional coefficient of each group of first bone points according to the matching score of each group of first bone points; and perform weighted sum operation according to the matching scores of multiple groups of first bone points and the proportional coefficient of each group of first bone points to obtain the first matching score.

[0295] Optionally, the calculation module 1104 is specifically used to determine the proportional coefficient of each group of second skeleton points according to the matching score of each group of second skeleton points; and perform weighted sum operation according to the matching scores of multiple groups of second skeleton points and the proportional coefficient of each group of second skeleton points to obtain the second matching score.

[0296] Optionally, the calculation module 1104 is specifically used to determine the proportional coefficient of each group of third skeletal points according to the matching score of each group of third skeletal points; and perform weighted sum operation according to the matching scores of multiple groups of third skeletal points and the proportional coefficient of each group of third skeletal points to obtain the third matching score.

[0297] Optionally, the calculation module 1104 is also used to determine multiple first weight coefficients of multiple groups of first skeleton points according to the matching scores of multiple groups of first skeleton points; determine multiple second weight coefficients of multiple groups of second skeleton points according to the matching scores of multiple groups of second skeleton points; determine multiple third weight coefficients of multiple groups of third skeleton points according to the third identification vectors of multiple groups of third skeleton points in the target skeleton points and the third standard vectors of the multiple groups of third skeleton points in the standard skeleton points; calculate the total matching score according to the first matching score, the second matching score and the third matching score; calculate the total weight coefficient according to the multiple first weight coefficients, the multiple second weight coefficients and the multiple third weight coefficients; calculate the target matching score according to the total matching score and the total weight coefficient.

[0298] Optionally, the calculation module 1104 is specifically configured to calculate a target matching score according to a ratio of a total matching score to a total weight coefficient.

[0299] Optionally, the action matching device 1100 may further include:

[0300] A judgment module, used to judge whether the target matching score is greater than or equal to a preset benchmark matching score;

[0301] The determination module is used to determine that the action to be matched has been completed and the preset action requirements corresponding to the action to be matched are met if the target matching score is greater than or equal to the benchmark matching score.

[0302] Optionally, the recognition module 1102 specifically adopts a target skeleton model to identify skeleton points of multiple frames of images to be matched with actions to be matched, and obtains multiple groups of first skeleton point positions, each group of first skeleton point positions includes: the positions of skeleton points identified from a frame of image to be matched; and the multiple groups of first skeleton point positions are averaged to obtain the target skeleton point position.

[0303] Optionally, the first acquisition module 1101 is further used to acquire a standard image with a preset standard action.

[0304] The recognition module 1102 is also used to use the target skeleton model to recognize the skeleton points of the standard image to obtain the positions of the standard skeleton points.

[0305] The action matching device 1100 may further include:

[0306] The storage module is used to store the standard skeleton point positions in the action template of the preset standard action.

[0307] Optionally, the recognition module 1102 is specifically used to use a target skeleton model to identify skeleton points of multiple frames of standard images with preset standard actions, and obtain multiple groups of second skeleton point positions, each group of second skeleton point positions includes: skeleton point positions identified from a frame of standard image; and averaging multiple groups of second skeleton point positions to obtain standard skeleton point positions.

[0308] The above-mentioned device is used to execute the method provided by the aforementioned embodiment, and its implementation principle and technical effect are similar, which will not be repeated here.

[0309] The above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more digital singnal processors (DSP), or one or more field programmable gate arrays (FPGA). For another example, when a module is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0310] Fig.12 A schematic diagram of a computer device provided in an embodiment of the present application, the computer device can be installed and run a terminal device or server with a motion matching application.

[0311] The computer device 1200 includes: a memory 1201 and a processor 1202. The memory 1201 and the processor 1202 are connected via a bus.

[0312] The memory 1201 is used to store programs, and the processor 1202 calls the programs stored in the memory 1201 to execute the above method embodiment. The specific implementation method and technical effect are similar and will not be repeated here.

[0313] Optionally, the present invention also provides a program product, such as a computer-readable storage medium, comprising a program, which is used to execute the above method embodiment when executed by a processor.

[0314] In the several embodiments provided by the present invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0315] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0316] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0317] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (English: processor) to perform some steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (English: Read-Only Memory, abbreviated: ROM), random access memory (English: Random Access Memory, abbreviated: RAM), disk or optical disk and other media that can store program codes.

[0318] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An action matching method, characterized in that: include: Acquire an image to be matched with an action to be matched; Using a target skeleton model corresponding to a preset standard action, the skeleton points of the image to be matched are identified to obtain the position of the target skeleton points; Acquire an action template of the preset standard action, wherein the action template of the preset standard action stores: a standard skeleton point position of the preset standard action; Calculating a target matching score between the action to be matched and the preset standard action according to the target skeleton point position and the standard skeleton point position, wherein the target matching score is used to characterize the matching degree between the action to be matched and the preset standard action; The step of calculating the target matching score between the action to be matched and the preset standard action according to the target skeleton point position and the standard skeleton point position includes: According to the positions of the multiple groups of first skeleton points in the target skeleton point positions and the positions of the multiple groups of first skeleton points in the standard skeleton point positions, the matching scores of the multiple groups of first skeleton points are calculated respectively; wherein the first skeleton points are: skeleton points with skeleton linkage in the human body; Calculating a first matching score according to the matching scores of the multiple groups of first skeleton points; According to the positions of the plurality of groups of second skeletal points in the target skeletal point positions and the positions of the plurality of groups of second skeletal points in the standard skeletal point positions, the matching scores of the plurality of groups of second skeletal points are calculated respectively; wherein the second skeletal points are skeletal points in the human body without skeletal linkage; Calculating a second matching score according to the matching scores of the plurality of groups of second skeleton points; According to the positions of the plurality of groups of third skeleton points in the target skeleton point positions and the positions of the plurality of groups of third skeleton points in the standard skeleton point positions, the matching scores of the plurality of groups of third skeleton points are calculated respectively; the third skeleton points are: skeleton points with non-fixed distances; Calculating a third matching score according to the matching scores of the plurality of groups of third skeleton points; The target matching score is calculated according to the first matching score, the second matching score and the third matching score.

2. The method according to claim 1, characterized in that The step of calculating the matching scores of the plurality of groups of first skeleton points respectively according to the positions of the plurality of groups of first skeleton points in the target skeleton point positions and the positions of the plurality of groups of first skeleton points in the standard skeleton point positions comprises: According to the first recognition vector formed by the position of each group of first skeleton points in the target skeleton point position and the first standard vector formed by the position of each group of first skeleton points in the standard skeleton point position, the matching score of each group of first skeleton points is calculated.

3. The method according to claim 2, characterized in that The step of calculating the matching score of each group of first skeleton points according to the first recognition vector formed by the position of each group of first skeleton points in the target skeleton point position and the first standard vector formed by the position of each group of first skeleton points in the standard skeleton point position comprises: An inner product operation is performed on the first recognition vector and the first standard vector to obtain a matching score for each group of first skeleton points.

4. The method according to claim 1, characterized in that The step of calculating the matching scores of the plurality of groups of second skeleton points respectively according to the positions of the plurality of groups of second skeleton points in the target skeleton point positions and the positions of the plurality of groups of second skeleton points in the standard skeleton point positions comprises: The matching score of each group of second skeleton points is calculated based on the second recognition vector formed by the position of each group of second skeleton points in the target skeleton point position and the second standard vector formed by the position of each group of second skeleton points in the standard skeleton point position.

5. The method according to claim 4, characterized in that The step of calculating the matching score of each group of second skeleton points according to the second recognition vector formed by the position of each group of second skeleton points in the target skeleton point position and the second standard vector formed by the position of each group of second skeleton points in the standard skeleton point position comprises: An inner product operation is performed on the second recognition vector and the second standard vector to obtain a matching score for each group of second skeleton points.

6. The method according to claim 1, characterized in that The step of calculating the matching scores of the plurality of groups of third skeleton points respectively according to the positions of the plurality of groups of third skeleton points in the target skeleton point positions and the positions of the plurality of groups of third skeleton points in the standard skeleton point positions comprises: The matching score of each group of third bone points is calculated based on the first bone distance between the positions of each group of third bone points in the target bone point position and the second bone distance between the positions of each group of third bone points in the standard bone point position.

7. The method according to claim 6, characterized in that The step of calculating the matching score of each group of third skeletal points according to the first skeletal distance between the positions of each group of third skeletal points in the target skeletal point positions and the second skeletal distance between the positions of each group of third skeletal points in the standard skeletal point positions comprises: Calculating a first relative distance of each group of third bone points according to the first bone distance and a first unit distance between the positions of the reference bones in the target bone point positions; Calculate the second relative distance of each group of third skeleton points according to the second skeleton distance and the second unit distance between the positions of the reference skeletons in the standard skeleton point positions; The matching score of each group of third skeleton points is calculated according to the first relative distance of each group of third skeleton points and the second relative distance of each group of third skeleton points.

8. The method according to claim 7, characterized in that The step of calculating the matching score of each group of third skeleton points according to the first relative distance of each group of third skeleton points and the second relative distance of each group of third skeleton points comprises: Calculating the first relative distance and the second relative distance, and calculating the bone distance deviation of each group of third bone points; The matching score of each group of third bone points is calculated according to the ratio of the bone distance deviation and the second relative distance.

9. The method according to claim 1, characterized in that: The calculating a first matching score according to the matching scores of the plurality of groups of first skeleton points comprises: Determining a proportionality coefficient of each group of first skeleton points according to the matching score of each group of first skeleton points; A weighted sum operation is performed according to the matching scores of the multiple groups of first skeleton points and the proportional coefficient of each group of first skeleton points to obtain the first matching score.

10. The method according to claim 1, characterized in that The calculating a second matching score according to the matching scores of the plurality of groups of second skeleton points comprises: Determining a proportional coefficient of each group of second skeleton points according to the matching score of each group of second skeleton points; A weighted sum operation is performed according to the matching scores of the multiple groups of second skeleton points and the proportional coefficient of each group of second skeleton points to obtain the second matching score.

11. The method according to claim 1, characterized in that: Calculating a third matching score according to the matching scores of the plurality of groups of third skeleton points includes: Determining a proportional coefficient of each group of third skeletal points according to the matching score of each group of third skeletal points; A weighted sum operation is performed according to the matching scores of the multiple groups of third skeleton points and the proportional coefficient of each group of third skeleton points to obtain the third matching score.

12. The method according to claim 1, characterized in that The method further comprises: Determining a plurality of first weight coefficients of the plurality of groups of first skeleton points respectively according to the matching scores of the plurality of groups of first skeleton points; Determining a plurality of second weight coefficients of the plurality of groups of second skeleton points respectively according to the matching scores of the plurality of groups of second skeleton points; Determine a plurality of third weight coefficients of the plurality of groups of third skeletal points respectively according to third identification vectors of the plurality of groups of third skeletal points in the target skeletal points and third standard vectors of the plurality of groups of third skeletal points in the standard skeletal points; The calculating the target matching score according to the first matching score, the second matching score and the third matching score includes: Calculating a total matching score according to the first matching score, the second matching score and the third matching score; Calculating a total weight coefficient according to the plurality of first weight coefficients, the plurality of second weight coefficients, and the plurality of third weight coefficients; The target matching score is calculated according to the total matching score and the total weight coefficient.

13. The method according to claim 12, characterized in that The calculating the target matching score according to the total matching score and the total weight coefficient includes: The target matching score is calculated according to the ratio of the total matching score to the total weight coefficient.

14. The method according to any one of claims 1 to 13, characterized in that: The method further comprises: Determining whether the target matching score is greater than or equal to a preset benchmark matching score; If the target matching score is greater than or equal to the reference matching score, it is determined that the action to be matched has been completed and the preset action requirement corresponding to the action to be matched is met.

15. The method according to any one of claims 1 to 13, characterized in that: The step of using a skeleton model corresponding to a preset standard action to identify the position of a target skeleton point in the image to be matched includes: Using the target skeleton model, identifying skeleton points of multiple frames of the to-be-matched images having the to-be-matched action, and obtaining multiple groups of first skeleton point positions, each group of first skeleton point positions including: positions of skeleton points identified from one frame of the to-be-matched image; The plurality of groups of first skeleton point positions are averaged to obtain the target skeleton point position.

16. The method according to any one of claims 1 to 13, characterized in that: Before obtaining the action template of the preset standard action, the method further includes: Acquire a standard image having the preset standard action; Using the target skeleton model, identifying the skeleton points of the standard image to obtain the positions of the standard skeleton points; The standard skeleton point position is stored in the action template of the preset standard action.

17. The method according to claim 16, characterized in that The method of using the target skeleton model to identify skeleton points of the standard image to obtain the positions of the standard skeleton points includes: Using the target skeleton model, identifying skeleton points of multiple frames of the standard image having the preset standard action, and obtaining multiple groups of second skeleton point positions, each group of second skeleton point positions includes: skeleton point positions identified from one frame of the standard image; The plurality of groups of second skeleton point positions are averaged to obtain the standard skeleton point position.

18. An action matching device, characterized in that: include: A first acquisition module, used for acquiring a to-be-matched image having an to-be-matched action; A recognition module is used to identify the skeleton points of the image to be matched by using the target skeleton model corresponding to the preset standard action to obtain the position of the target skeleton point; A second acquisition module is used to acquire an action template of the preset standard action, wherein the action template of the preset standard action stores: a standard skeleton point position of the preset standard action; A calculation module, used for calculating a target matching score between the action to be matched and the preset standard action according to the target skeleton point position and the standard skeleton point position, wherein the target matching score is used to characterize the matching degree between the action to be matched and the preset standard action; The calculation module is used to calculate the matching scores of the multiple groups of first skeleton points according to the positions of the multiple groups of first skeleton points in the target skeleton point positions and the positions of the multiple groups of first skeleton points in the standard skeleton point positions; wherein the first skeleton points are: skeleton points with skeleton linkage in the human body; Calculating a first matching score according to the matching scores of the multiple groups of first skeleton points; According to the positions of the plurality of groups of second skeletal points in the target skeletal point positions and the positions of the plurality of groups of second skeletal points in the standard skeletal point positions, the matching scores of the plurality of groups of second skeletal points are calculated respectively; wherein the second skeletal points are skeletal points in the human body without skeletal linkage; Calculating a second matching score according to the matching scores of the plurality of groups of second skeleton points; According to the positions of the plurality of groups of third skeleton points in the target skeleton point positions and the positions of the plurality of groups of third skeleton points in the standard skeleton point positions, the matching scores of the plurality of groups of third skeleton points are calculated respectively; the third skeleton points are: skeleton points with non-fixed distances; Calculating a third matching score according to the matching scores of the plurality of groups of third skeleton points; The target matching score is calculated according to the first matching score, the second matching score and the third matching score.

19. A computer device, characterized in that: include: A memory and a processor, wherein the memory stores a computer program executable by the processor, and the processor implements the action matching method described in any one of claims 1 to 17 when executing the computer program.

20. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is read and executed, the action matching method described in any one of claims 1 to 17 is implemented.

Images