Motion posture evaluation system and method based on image processing

Abstract

The invention discloses a motion posture evaluation system and method based on image processing, and relates to the technical field of motion posture evaluation, and the method comprises the steps: obtaining a motion video of a student, evaluating the motion approximation between the motion video and a teaching video, and obtaining a reference teaching video; obtaining a reference teaching video of the motion video, generating a reference image set of the reference teaching video, and evaluating the attitude abnormality degree of attitude image frames in the attitude image set to obtain abnormal attitude image frames; acquiring an abnormal attitude image frame in the attitude image set, acquiring position data of a motion key point in the abnormal attitude image frame, and evaluating attitude abnormality of the motion key point in the abnormal attitude image frame in combination with the reference image set to obtain motion attitude abnormal data; and acquiring motion posture abnormal data of the motion video, generating motion posture adjustment data of the student according to the motion posture abnormal data, and performing intelligent adjustment on the motion posture of the student.

Description

Technical Field

[0001] The present invention relates to the technical field of motion posture evaluation, and in particular to a motion posture evaluation system and method based on image processing. Background Art

[0002] At present, the assessment of the posture of trainees during exercise mainly includes visual observation method and video analysis method. The visual observation method is to directly observe the movements and postures of trainees through the eyes of the assessor, but this method is highly subjective and depends heavily on the experience and skills of the assessor. The video analysis method uses a shooting device to shoot the movements of trainees, and then the assessor plays back and analyzes the shot video. Although this method can make subtle observations many times and overcomes the limitations of direct observation with the naked eye, it also requires assessors with professional knowledge to analyze the video, and it will also cause errors in analysis due to human subjectivity. Therefore, the current methods for assessing the posture of trainees during exercise basically require assessors to directly or indirectly assess the trainees. These assessment methods not only require a large number of talents with professional knowledge and skills, which consumes a lot of manpower and financial resources, but also rely solely on manual analysis, which can easily lead to large errors between the assessment results of the trainees and the actual situation due to subjectivity. Summary of the invention

[0003] The purpose of the present invention is to provide a motion posture evaluation system and method based on image processing to solve the problems raised in the prior art.

[0004] To achieve the above object, the present invention provides the following technical solution: a motion posture evaluation method based on image processing, the method comprising:

[0005] Step S100: constructing a posture evaluation cloud platform, obtaining a student's motion video, obtaining a teaching video in the cloud platform, evaluating the motion similarity between the motion video and the teaching video, and obtaining a reference teaching video;

[0006] Step S200: obtaining a posture image set of a trainee, obtaining a reference teaching video of a sports video, generating a reference image set of the reference teaching video, evaluating the posture abnormality of the posture image frames in the posture image set, and obtaining abnormal posture image frames;

[0007] Step S300: Acquire abnormal posture image frames in the posture image set, acquire position data of motion key points in the abnormal posture image frames, and evaluate the posture abnormality of the motion key points in the abnormal posture image frames in combination with the reference image set to obtain motion posture abnormality data;

[0008] Step S400: Acquire abnormal motion posture data of the motion video, and generate motion posture adjustment data of the trainee according to the abnormal motion posture data, and intelligently adjust the motion posture of the trainee.

[0009] Furthermore, step S100 includes:

[0010] Step S101: When the trainee uploads the filmed sports video to the cloud platform in the program, the sports video is acquired, and the key frames in the sports video are extracted at every characteristic duration according to the preset characteristic duration to obtain posture image frames, and the posture image frames are preprocessed, and each posture image frame of the sports video is acquired and collected to obtain a posture image set of the sports video;

[0011] Step S102: randomly selecting a number of posture image frames from the posture image set and recording them as marked posture image frames, acquiring and collecting a number of marked posture image frames in the posture image set to obtain a motion recognition set of the motion video;

[0012] Step S103: Obtain preset key points of various motions, establish a two-dimensional coordinate system for the marked posture image frame, obtain the coordinates of the key points of various motions in the marked posture image frame, and obtain the feature vector F of the marked posture image frame:

[0013]

[0014] Where x1 and y1 represent the horizontal and vertical coordinates of the first key point of the motion in the marked posture image frame, respectively; x2 and y2 represent the horizontal and vertical coordinates of the second key point of the motion in the marked posture image frame, respectively; m ,y m They are respectively represented as the horizontal and vertical coordinates of the mth motion key point in the marked posture image frame; m represents the total number of various motion key points; W represents the image width of the marked posture image frame; H represents the image width of the marked posture image frame;

[0015] Step S104: Calculate the feature aggregation vector F of the motion recognition set avg :

[0016]

[0017] Where ε represents the total number of labeled posture image frames in the motion recognition set; F i It is represented as the feature vector of the i-th labeled pose image frame in the motion recognition set;

[0018] Step S105: Obtain several standard posture images of each teaching video from the cloud platform, obtain the feature vector of the standard posture image, obtain the motion recognition set, and the motion approximation value of each teaching video, wherein the specific acquisition process of obtaining the motion recognition set and the motion approximation value of the a-th teaching video in the cloud platform is:

[0019] Calculate the motion approximation value of the motion recognition set and several standard posture images of the a-th teaching video, where the motion approximation value R of the motion recognition set and the b-th standard posture image of the a-th teaching video is a,b :

[0020]

[0021] Among them, D a,b Represented as the feature vector of the bth standard posture image;

[0022] Step S106: Calculate the motion recognition set and the motion approximation C of the a-th teaching video a :

[0023]

[0024] Where j represents the total number of standard posture images in the a-th teaching video; R a,z It is represented as a motion recognition set, and the feature motion approximation of the zth standard posture image of the ath teaching video;

[0025] Step S107: Evaluate the motion video and perform motion approximation with each teaching video in the cloud platform. The specific process is to obtain the motion recognition set and the maximum value of the motion approximation value of each teaching video, obtain the teaching video corresponding to the maximum value, and determine that the motion video has the greatest motion approximation with the teaching video in the cloud platform, and record the teaching video as the reference teaching video of the motion video.

[0026] Further, step S200 includes:

[0027] Step S201: obtaining a reference teaching video of a sports video, extracting key frames of the sports video at every characteristic duration to obtain reference posture image frames, and collecting several reference posture image frames of the reference teaching video to obtain a reference image set of the reference teaching video;

[0028] Step S202: Acquire feature vectors of several reference posture image frames, calculate the posture approximation between each posture image frame in the posture image set of the motion video and several reference posture image frames in the reference image set, wherein the posture approximation H between the αth posture image frame in the posture image set and the βth reference posture image frame in the reference image set is α,β :

[0029]

[0030] Among them, G α Represented as the feature vector of the αth pose image frame; Q β Represented as the feature vector of the βth reference pose image frame;

[0031] Step S203: obtaining the maximum value of the posture approximation value between the αth posture image frame in the posture image set and a plurality of reference posture image frames, and taking the reference posture image frame corresponding to the maximum value of the posture approximation value as the target reference posture image frame of the αth posture image frame;

[0032] Step S204: evaluating the abnormality of the posture of the αth posture image frame in the posture image set. The specific evaluation process is that when the posture approximation value of the αth posture image frame and the target reference posture image frame of the αth posture image frame is greater than or equal to a preset posture approximation threshold, it is determined that the movement posture of the trainee in the αth posture image frame is not abnormal;

[0033] Step S205: When the posture approximation value of the αth posture image frame and the target reference posture image frame of the αth posture image frame is less than a preset posture approximation threshold, it is determined that the movement posture of the student in the αth posture image frame is abnormal, and the αth posture image frame is recorded as an abnormal posture image frame, and several abnormal posture image frames in the posture image set are acquired.

[0034] Furthermore, step S300 includes:

[0035] Step S301: acquiring a number of abnormal posture image frames in the posture image set, acquiring target reference posture image frames corresponding to the abnormal posture image frames, establishing two-dimensional coordinate systems for the target reference posture image frames of the abnormal posture image frames, and acquiring position data of the trainee's motion key points from the abnormal posture image frames, the position data including the coordinate vectors of each motion key point in the abnormal posture image frames;

[0036] Step S302: obtaining motion key points of joint points of several preset characteristic joints in the reference teaching video to which the target reference posture image frame belongs, and recording them as characteristic motion key points of the characteristic joints, obtaining adjacent motion key points of the characteristic motion key points, and recording them as first motion key points and second motion key points of the characteristic joints respectively;

[0037] Step S303: Evaluate the abnormality of the posture of the motion key points in the abnormal posture image frame. The specific evaluation process is: calculate the characteristic joint angles of several characteristic joints in the abnormal posture image frame, wherein the characteristic joint angle θ of the δth characteristic joint in the abnormal posture image frame is δ:

[0038]

[0039] Among them, S δ It is represented as the coordinate vector of the key point of the characteristic motion of the δth characteristic joint; S δ 1 It is represented as the coordinate vector of the first motion key point of the δth feature joint; S δ 2 It is represented as the coordinate vector of the second motion key point of the δth feature joint;

[0040] Step S304: Calculate the angle deviation value △θ between the δth feature joint in the abnormal posture image frame and the feature joint angle deviation value △θ of the target reference posture image frame δ =θ δ ′-θ δ , when the characteristic joint angle deviation value △θ δ , outside the preset angle deviation range, determining that the posture of the δth feature joint in the abnormal posture image frame is abnormal, and taking the δth feature joint as the feature abnormal joint in the abnormal posture image frame;

[0041] Step S305: acquiring and collecting several characteristic abnormal joints of the abnormal posture image frame to obtain abnormal motion posture data of the trainee in the abnormal posture image frame;

[0042] The above steps first obtain the motion key points of the joint points of several preset feature joints, and then calculate the feature joints in the abnormal posture image frame and the feature joint angle deviation value of the target reference posture image frame. It is possible to know the area where the feature joints are located and whether the student's movements are standard, which helps to quickly find the area where the student's movement posture is wrong, thereby better evaluating the student's movement posture.

[0043] Furthermore, step S400 includes:

[0044] Step S401: obtaining a plurality of abnormal posture image frames in a motion video, obtaining motion posture abnormality data of the plurality of abnormal posture image frames, and extracting characteristic joint angle deviation values ​​of characteristic abnormal joints of the abnormal posture image frames from the motion posture abnormality data;

[0045] Step S402: Mark several characteristic abnormal joints in several abnormal posture image frames and corresponding target reference posture image frames respectively, generate the trainee's motion posture adjustment data, and display it to the trainee in the program.

[0046] In order to better implement the above method, a motion posture evaluation system based on image processing is also proposed, which includes a motion posture approximate evaluation module, an image posture evaluation module, a motion posture abnormal data module, and a posture adjustment module;

[0047] A motion posture approximation evaluation module is used to evaluate the motion approximation between the motion video and the teaching video to obtain a reference teaching video;

[0048] An image posture evaluation module is used to evaluate the posture abnormality of the posture image frames in the posture image set to obtain abnormal posture image frames;

[0049] The motion posture abnormality data module is used to obtain the position data of the motion key points in the abnormal posture image frame, and evaluate the posture abnormality of the motion key points in the abnormal posture image frame in combination with the reference image set to obtain the motion posture abnormality data;

[0050] The posture adjustment module is used to obtain abnormal motion posture data of the motion video, and generate motion posture adjustment data of the trainee based on the abnormal motion posture data, so as to intelligently adjust the motion posture of the trainee.

[0051] Further, the motion posture approximation evaluation module includes a motion approximation value unit and a motion posture approximation evaluation unit;

[0052] A motion approximation unit is used to calculate the motion approximation of the motion recognition set and each teaching video in the cloud platform;

[0053] The motion posture approximation evaluation unit is used to evaluate the motion approximation between the motion video and each teaching video in the cloud platform to obtain a reference teaching video in the motion video.

[0054] Further, the image posture evaluation module includes a posture approximation unit and an image posture evaluation unit;

[0055] A posture approximation unit, used for calculating the posture approximation between each posture image frame in the posture image set of the motion video and a plurality of reference posture image frames in the reference image set;

[0056] The image posture evaluation unit is used to evaluate the difference between the αth posture image frame in the posture image set and the motion posture in the reference image set to obtain an abnormal posture image frame.

[0057] Further, the motion posture abnormal data module includes a characteristic joint angle unit and a motion posture abnormal data unit;

[0058] A characteristic joint angle unit is used to calculate characteristic joint angles of several characteristic joints in an abnormal posture image frame;

[0059] The motion posture abnormal data unit is used to acquire and collect several characteristic abnormal joints of the abnormal posture image frame to obtain the motion posture abnormal data.

[0060] Further, the posture adjustment module includes a posture adjustment unit;

[0061] The posture adjustment unit is used to obtain several abnormal posture image frames in the motion video, obtain the motion posture abnormality data of several abnormal posture image frames, mark several characteristic abnormal joints in several abnormal posture image frames and corresponding target reference posture image frames, and generate the trainee's motion posture adjustment data, and show it to the trainee in the program.

[0062] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention evaluates the abnormal posture in the student's motion video, first by evaluating the motion similarity between the motion video and the teaching video, automatically finding the reference teaching video corresponding to the motion video, and evaluating the degree of posture abnormality of the posture image frame in the posture image set, thereby finding the abnormal posture image frame with abnormality, and analyzing the angle of the motion key point of the characteristic joint in the abnormal posture image frame, so as to determine which characteristic joint of the student has a posture problem in the abnormal posture image frame, and generate visual adjustment data, and intelligently adjust the student's motion posture, which not only avoids the errors caused by the subjectivity of manual analysis, but also greatly saves manpower and material resources, and greatly improves the analysis speed and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a method flow chart of a system and method for evaluating a motion posture based on image processing of the present invention;

[0064] Figure 2 It is a module schematic diagram of a motion posture evaluation system and method based on image processing of the present invention. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] Example: Figure 1-Figure 2 As shown, the present invention provides a technical solution, a motion posture evaluation method based on image processing, the method comprising:

[0067] Step S100: constructing a posture evaluation cloud platform, obtaining a student's motion video, obtaining a teaching video in the cloud platform, evaluating the motion similarity between the motion video and the teaching video, and obtaining a reference teaching video;

[0068] Wherein, step S100 includes:

[0069] Step S101: When the trainee uploads the filmed sports video to the cloud platform in the program, the sports video is acquired, and the key frames in the sports video are extracted at every characteristic duration according to the preset characteristic duration to obtain posture image frames, and the posture image frames are preprocessed, and each posture image frame of the sports video is acquired and collected to obtain a posture image set of the sports video;

[0070] For example, preprocessing generally includes processing steps such as resizing and normalization;

[0071] Step S102: randomly selecting a number of posture image frames from the posture image set and recording them as marked posture image frames, acquiring and collecting a number of marked posture image frames in the posture image set to obtain a motion recognition set of the motion video;

[0072] Step S103: Obtain preset key points of various motions, establish a two-dimensional coordinate system for the marked posture image frame, obtain the coordinates of the key points of various motions in the marked posture image frame, and obtain the feature vector F of the marked posture image frame:

[0073]

[0074] Where x1 and y1 represent the horizontal and vertical coordinates of the first key point of the motion in the marked posture image frame, respectively; x2 and y2 represent the horizontal and vertical coordinates of the second key point of the motion in the marked posture image frame, respectively; m ,y m They are respectively represented as the horizontal and vertical coordinates of the mth motion key point in the marked posture image frame; m represents the total number of various motion key points; W represents the image width of the marked posture image frame; H represents the image width of the marked posture image frame;

[0075] For example, the key points of various movements include the wrist, left shoulder, etc.;

[0076] Step S104: Calculate the feature aggregation vector F of the motion recognition set avg :

[0077]

[0078] Where ε represents the total number of labeled posture image frames in the motion recognition set; F i It is represented as the feature vector of the i-th labeled pose image frame in the motion recognition set;

[0079] Step S105: Obtain several standard posture images of each teaching video from the cloud platform, obtain the feature vector of the standard posture image, obtain the motion recognition set, and the motion approximation value of each teaching video, wherein the specific acquisition process of obtaining the motion recognition set and the motion approximation value of the a-th teaching video in the cloud platform is:

[0080] Calculate the motion approximation value of the motion recognition set and several standard posture images of the a-th teaching video, where the motion approximation value R of the motion recognition set and the b-th standard posture image of the a-th teaching video is a,b :

[0081]

[0082] Among them, D a,b Represented as the feature vector of the bth standard posture image;

[0083] Step S106: Calculate the motion recognition set and the motion approximation C of the a-th teaching video a :

[0084]

[0085] Where j represents the total number of standard posture images in the a-th teaching video; R a,z It is represented as a motion recognition set, and the feature motion approximation of the zth standard posture image of the ath teaching video;

[0086] For example, the total number of standard posture images of the first teaching video is 5; the motion recognition set, and the feature motion approximation value R of the first standard posture image of the first teaching video 1,1 Expressed as 0.8; motion recognition set, feature motion approximation R of the second standard posture image of the first teaching video 1,2 Expressed as 0.7; motion recognition set, feature motion approximation R of the third standard posture image of the first teaching video 1,3 Expressed as 0.9; motion recognition set, feature motion approximation R of the fourth standard posture image of the first teaching video 1,4 Expressed as 0.5; motion recognition set, feature motion approximation R of the fifth standard posture image of the first teaching video 1,5 It is expressed as 0.8;

[0087] Calculate the motion recognition set and the motion approximation C1 of the first teaching video:

[0088]

[0089] Step S107: Evaluate the motion approximation of the motion video with each teaching video in the cloud platform. The specific process is to obtain the maximum value of the motion approximation value of the motion recognition set and each teaching video, obtain the teaching video corresponding to the maximum value, and determine that the motion approximation of the motion video with the teaching video in the cloud platform is the greatest, and record the teaching video as the reference teaching video of the motion video;

[0090] Step S200: obtaining a posture image set of a trainee, obtaining a reference teaching video of a sports video, generating a reference image set of the reference teaching video, evaluating the posture abnormality of the posture image frames in the posture image set, and obtaining abnormal posture image frames;

[0091] Wherein, step S200 includes:

[0092] Step S201: obtaining a reference teaching video of a sports video, extracting key frames of the sports video at every characteristic duration to obtain reference posture image frames, and collecting several reference posture image frames of the reference teaching video to obtain a reference image set of the reference teaching video;

[0093] Step S202: Acquire feature vectors of several reference posture image frames, calculate the posture approximation between each posture image frame in the posture image set of the motion video and several reference posture image frames in the reference image set, wherein the posture approximation H between the αth posture image frame in the posture image set and the βth reference posture image frame in the reference image set is α,β :

[0094]

[0095] Among them, G α Represented as the feature vector of the αth pose image frame; Q β Represented as the feature vector of the βth reference pose image frame;

[0096] Step S203: obtaining the maximum value of the posture approximation value between the αth posture image frame in the posture image set and a plurality of reference posture image frames, and taking the reference posture image frame corresponding to the maximum value of the posture approximation value as the target reference posture image frame of the αth posture image frame;

[0097] Step S204: evaluating the abnormality of the posture of the αth posture image frame in the posture image set. The specific evaluation process is that when the posture approximation value of the αth posture image frame and the target reference posture image frame of the αth posture image frame is greater than or equal to a preset posture approximation threshold, it is determined that the movement posture of the trainee in the αth posture image frame is not abnormal;

[0098] Step S205: when the posture approximation value of the αth posture image frame and the target reference posture image frame of the αth posture image frame is less than a preset posture approximation threshold, it is determined that the movement posture of the trainee in the αth posture image frame is abnormal, and the αth posture image frame is recorded as an abnormal posture image frame, and a plurality of abnormal posture image frames in the posture image set are acquired;

[0099] Step S300: Acquire abnormal posture image frames in the posture image set, acquire position data of motion key points in the abnormal posture image frames, and evaluate the posture abnormality of the motion key points in the abnormal posture image frames in combination with the reference image set to obtain motion posture abnormality data;

[0100] Wherein, step S300 includes:

[0101] Step S301: acquiring a number of abnormal posture image frames in the posture image set, acquiring target reference posture image frames corresponding to the abnormal posture image frames, establishing two-dimensional coordinate systems for the target reference posture image frames of the abnormal posture image frames, and acquiring position data of the trainee's motion key points from the abnormal posture image frames, the position data including the coordinate vectors of each motion key point in the abnormal posture image frames;

[0102] Step S302: obtaining motion key points of joint points of several preset characteristic joints in the reference teaching video to which the target reference posture image frame belongs, and recording them as characteristic motion key points of the characteristic joints, obtaining adjacent motion key points of the characteristic motion key points, and recording them as first motion key points and second motion key points of the characteristic joints respectively;

[0103] For example, several characteristic joints include shoulder joint, elbow joint, knee joint, etc.;

[0104] Step S303: Evaluate the abnormality of the posture of the motion key points in the abnormal posture image frame. The specific evaluation process is: calculate the characteristic joint angles of several characteristic joints in the abnormal posture image frame, wherein the characteristic joint angle θ of the δth characteristic joint in the abnormal posture image frame is δ :

[0105]

[0106] Among them, S δ It is represented as the coordinate vector of the key point of the characteristic motion of the δth characteristic joint; S δ 1 It is represented as the coordinate vector of the first motion key point of the δth feature joint; S δ 2 It is represented as the coordinate vector of the second motion key point of the δth feature joint;

[0107] Step S304: Calculate the angle deviation value △θ between the δth feature joint in the abnormal posture image frame and the feature joint angle deviation value △θ of the target reference posture image frame δ =θ δ ′-θ δ , when the characteristic joint angle deviation value △θ δ , outside the preset angle deviation range, determining that the posture of the δth feature joint in the abnormal posture image frame is abnormal, and taking the δth feature joint as the feature abnormal joint in the abnormal posture image frame;

[0108] Step S305: acquiring and collecting several characteristic abnormal joints of the abnormal posture image frame to obtain abnormal motion posture data of the trainee in the abnormal posture image frame;

[0109] Step S400: Acquire abnormal motion posture data of the motion video, and generate motion posture adjustment data of the trainee according to the abnormal motion posture data, and intelligently adjust the motion posture of the trainee;

[0110] Wherein, step S400 includes:

[0111] Step S401: obtaining a plurality of abnormal posture image frames in a motion video, obtaining motion posture abnormality data of the plurality of abnormal posture image frames, and extracting characteristic joint angle deviation values ​​of characteristic abnormal joints of the abnormal posture image frames from the motion posture abnormality data;

[0112] Step S402: marking several characteristic abnormal joints in several abnormal posture image frames and corresponding target reference posture image frames, and generating the trainee's motion posture adjustment data, which is displayed to the trainee in the program;

[0113] In order to better implement the above method, a motion posture evaluation system based on image processing is also proposed, which includes a motion posture approximate evaluation module, an image posture evaluation module, a motion posture abnormal data module, and a posture adjustment module;

[0114] A motion posture approximation evaluation module is used to evaluate the motion approximation between the motion video and the teaching video to obtain a reference teaching video;

[0115] An image posture evaluation module is used to evaluate the posture abnormality of the posture image frames in the posture image set to obtain abnormal posture image frames;

[0116] The motion posture abnormality data module is used to obtain the position data of the motion key points in the abnormal posture image frame, and evaluate the posture abnormality of the motion key points in the abnormal posture image frame in combination with the reference image set to obtain the motion posture abnormality data;

[0117] The posture adjustment module is used to obtain abnormal motion posture data of the motion video, and generate motion posture adjustment data of the trainee according to the abnormal motion posture data, so as to intelligently adjust the motion posture of the trainee;

[0118] Wherein, the motion posture approximation evaluation module includes a motion approximation value unit and a motion posture approximation evaluation unit;

[0119] A motion approximation unit is used to calculate the motion approximation of the motion recognition set and each teaching video in the cloud platform;

[0120] A motion posture approximation evaluation unit is used to evaluate the motion approximation of the motion video and each teaching video in the cloud platform to obtain a reference teaching video in the motion video;

[0121] Wherein, the image posture evaluation module includes a posture approximation unit and an image posture evaluation unit;

[0122] A posture approximation unit, used for calculating the posture approximation between each posture image frame in the posture image set of the motion video and a plurality of reference posture image frames in the reference image set;

[0123] An image posture evaluation unit is used to evaluate the difference between the αth posture image frame in the posture image set and the motion posture in the reference image set to obtain an abnormal posture image frame;

[0124] Among them, the motion posture abnormal data module includes a characteristic joint angle unit and a motion posture abnormal data unit;

[0125] A characteristic joint angle unit is used to calculate characteristic joint angles of several characteristic joints in an abnormal posture image frame;

[0126] A motion posture abnormal data unit is used to acquire and collect several characteristic abnormal joints of the abnormal posture image frame to obtain motion posture abnormal data;

[0127] Wherein, the posture adjustment module includes a posture adjustment unit;

[0128] The posture adjustment unit is used to obtain several abnormal posture image frames in the motion video, obtain the motion posture abnormality data of several abnormal posture image frames, mark several characteristic abnormal joints in several abnormal posture image frames and corresponding target reference posture image frames, and generate the trainee's motion posture adjustment data, and show it to the trainee in the program.

[0129] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A motion posture assessment method based on image processing, characterized in that: The method comprises: Step S100: constructing a posture evaluation cloud platform, obtaining a trainee's motion video, obtaining a teaching video in the cloud platform, evaluating the motion similarity between the motion video and the teaching video, and obtaining a reference teaching video; Step S200: obtaining a posture image set of the trainee, obtaining a reference teaching video of the sports video, generating a reference image set of the reference teaching video, evaluating the posture abnormality of the posture image frames in the posture image set, and obtaining abnormal posture image frames; Step S300: Acquire abnormal posture image frames in the posture image set, acquire position data of motion key points in the abnormal posture image frames, and evaluate the posture abnormality of the motion key points in the abnormal posture image frames in combination with the reference image set to obtain motion posture abnormality data; Step S400: Acquire abnormal motion posture data of the motion video, and generate motion posture adjustment data of the trainee according to the abnormal motion posture data, so as to intelligently adjust the motion posture of the trainee.

2. The motion posture evaluation method based on image processing according to claim 1, characterized in that: The step S100 includes: Step S101: When a student uploads a filmed sports video to the cloud platform in the program, the sports video is acquired, and key frames in the sports video are extracted at every characteristic duration according to a preset characteristic duration to obtain posture image frames, and the posture image frames are preprocessed, and each posture image frame of the sports video is acquired and collected to obtain a posture image set of the sports video; Step S102: randomly selecting a number of gesture image frames from the gesture image set and recording them as marked gesture image frames, acquiring and collecting the number of marked gesture image frames in the gesture image set to obtain a motion recognition set of the motion video; Step S103: Acquire various preset key points of motion, establish a two-dimensional coordinate system for the marked posture image frame, acquire the coordinates of various key points of motion in the marked posture image frame, and acquire the feature vector F of the marked posture image frame: Wherein, x1 and y1 represent the horizontal coordinate and vertical coordinate of the first key point of motion in the marked posture image frame, respectively; x2 and y2 represent the horizontal coordinate and vertical coordinate of the second key point of motion in the marked posture image frame, respectively; m ,y m They are respectively represented as the horizontal coordinate and the vertical coordinate of the mth motion key point in the marked posture image frame; m represents the total number of the motion key points; W represents the image width of the marked posture image frame; H represents the image width of the marked posture image frame; Step S104: Calculate the feature aggregation vector F of the motion recognition set avg : Wherein, ε represents the total number of image frames with marked postures in the motion recognition set; F i is represented as the feature vector of the i-th labeled posture image frame in the motion recognition set; Step S105: obtaining several standard posture images of each teaching video from the cloud platform, obtaining feature vectors of the standard posture images, obtaining the motion recognition set, and the motion approximation values ​​of each teaching video, wherein the specific process of obtaining the motion recognition set and the motion approximation value of the a-th teaching video in the cloud platform is as follows: Calculate the characteristic motion approximation value of the motion recognition set and several standard posture images of the a-th teaching video, wherein the characteristic motion approximation value R of the motion recognition set and the b-th standard posture image of the a-th teaching video is a,b : Among them, D a,b is represented as the feature vector of the bth standard posture image; Step S106: Calculate the motion recognition set and the motion approximation C of the a-th teaching video a : Wherein, j represents the total number of standard posture images of the a-th teaching video; R a,z It is represented as the motion recognition set and the characteristic motion approximation value of the zth standard posture image of the ath teaching video; Step S107: Evaluate the motion approximation of the motion video with the various teaching videos in the cloud platform. The specific process is to obtain the maximum value of the motion approximation value of the motion recognition set and the various teaching videos, obtain the teaching video corresponding to the maximum value, and determine that the motion video has the greatest motion approximation with the teaching videos in the cloud platform, and record the teaching video as the reference teaching video of the motion video.

3. A motion posture assessment method based on image processing according to claim 2, characterized in that: The step S200 includes: Step S201: obtaining a reference teaching video of the sports video, extracting key frames of the sports video at every characteristic duration to obtain reference posture image frames, and collecting a plurality of reference posture image frames of the reference teaching video to obtain a reference image set of the reference teaching video; Step S202: Acquire feature vectors of the plurality of reference posture image frames, and calculate posture approximations between each posture image frame in the posture image set of the motion video and the plurality of reference posture image frames in the reference image set, wherein the posture approximation H between the αth posture image frame in the posture image set and the βth reference posture image frame in the reference image set is α,β : Among them, G α It is represented as the feature vector of the αth posture image frame; Q β Represented as the feature vector of the βth reference pose image frame; Step S203: obtaining the maximum value of the posture approximation values ​​between the αth posture image frame in the posture image set and the plurality of reference posture image frames, and taking the reference posture image frame corresponding to the maximum value of the posture approximation value as the target reference posture image frame of the αth posture image frame; Step S204: evaluating the abnormality of the posture of the αth posture image frame in the posture image set, wherein the specific evaluation process is that when the posture approximation value of the αth posture image frame and the target reference posture image frame of the αth posture image frame is greater than or equal to a preset posture approximation threshold, it is determined that the movement posture of the trainee in the αth posture image frame is not abnormal; Step S205: When the posture approximation value of the αth posture image frame and the target reference posture image frame of the αth posture image frame is less than a preset posture approximation threshold, it is determined that the movement posture of the student in the αth posture image frame is abnormal, and the αth posture image frame is recorded as an abnormal posture image frame, and several abnormal posture image frames in the posture image set are acquired.

4. The method for evaluating motion posture based on image processing according to claim 3, characterized in that: The step S300 includes: Step S301: acquiring a plurality of abnormal posture image frames in the posture image set, acquiring target reference posture image frames corresponding to the abnormal posture image frames, establishing two-dimensional coordinate systems for the target reference posture image frames of the abnormal posture image frames respectively, and acquiring position data of the motion key points of the trainee from the abnormal posture image frames, wherein the position data includes coordinate vectors of each motion key point in the abnormal posture image frames; Step S302: obtaining motion key points of joint points of several preset characteristic joints in the reference teaching video to which the target reference posture image frame belongs, and recording them as characteristic motion key points of the characteristic joints, obtaining adjacent motion key points of the characteristic motion key points, and recording them as first motion key points and second motion key points of the characteristic joints respectively; Step S303: Evaluate the abnormality of the posture of the motion key points in the abnormal posture image frame. The specific evaluation process is: calculate the characteristic joint angles of several characteristic joints in the abnormal posture image frame, wherein the characteristic joint angle θ of the δth characteristic joint in the abnormal posture image frame is δ : Among them, S δ S is represented as the coordinate vector of the key point of the characteristic motion of the δth characteristic joint; δ 1 S is represented as the coordinate vector of the first motion key point of the δth feature joint; δ 2 The coordinate vector of the second motion key point of the δth feature joint is represented as; Step S304: Calculate the angle deviation value Δθ between the δth feature joint in the abnormal posture image frame and the feature joint angle deviation value Δθ of the target reference posture image frame δ =θ δ ′-θ δ , when the characteristic joint angle deviation value △θ δ , outside the preset angle deviation range, determining that the posture of the δth feature joint in the abnormal posture image frame is abnormal, and using the δth feature joint as the feature abnormal joint in the abnormal posture image frame; Step S305: Acquire and aggregate several characteristic abnormal joints of the abnormal posture image frame to obtain the abnormal motion posture data of the trainee in the abnormal posture image frame.

5. The motion posture assessment method based on image processing according to claim 4, characterized in that: The step S400 includes: Step S401: obtaining a plurality of abnormal posture image frames in the motion video, obtaining motion posture abnormality data of the plurality of abnormal posture image frames, and extracting characteristic joint angle deviation values ​​of characteristic abnormal joints of the abnormal posture image frames from the motion posture abnormality data respectively; Step S402: marking the plurality of characteristic abnormal joints in the plurality of abnormal posture image frames and the corresponding target reference posture image frames respectively, and generating the motion posture adjustment data of the trainee, and displaying it to the trainee in the program.

6. A motion posture assessment system based on image processing, used to execute a motion posture assessment method based on image processing according to any one of claims 1 to 5, characterized in that: The system includes a motion posture approximation assessment module, an image posture assessment module, a motion posture abnormal data module, and a posture adjustment module; The motion posture approximation evaluation module is used to evaluate the motion approximation between the motion video and the teaching video to obtain a reference teaching video; The image posture evaluation module is used to evaluate the posture abnormality degree of the posture image frames in the posture image set to obtain abnormal posture image frames; The motion posture abnormality data module is used to obtain the position data of the motion key points in the abnormal posture image frame, and evaluate the posture abnormality of the motion key points in the abnormal posture image frame in combination with the reference image set to obtain the motion posture abnormality data; The posture adjustment module is used to obtain the abnormal motion posture data of the motion video, and generate the motion posture adjustment data of the trainee according to the abnormal motion posture data, so as to intelligently adjust the motion posture of the trainee.

7. The motion posture assessment system based on image processing according to claim 6, characterized in that: The motion posture approximation evaluation module includes a motion approximation value unit and a motion posture approximation evaluation unit; The motion approximation unit is used to calculate the motion approximation of the motion recognition set and each teaching video in the cloud platform; The motion posture approximation evaluation unit is used to evaluate the motion approximation between the motion video and each teaching video in the cloud platform to obtain a reference teaching video in the motion video.

8. The motion posture assessment system based on image processing according to claim 6, characterized in that: The image posture evaluation module includes a posture approximation unit and an image posture evaluation unit; The posture approximation unit is used to calculate the posture approximation between each posture image frame in the posture image set of the motion video and the plurality of reference posture image frames in the reference image set; The image posture evaluation unit is used to evaluate the difference between the αth posture image frame in the posture image set and the motion posture in the reference image set to obtain an abnormal posture image frame.

9. The motion posture assessment system based on image processing according to claim 6, characterized in that: The motion posture abnormal data module includes a characteristic joint angle unit and a motion posture abnormal data unit; The characteristic joint angle unit is used to calculate the characteristic joint angles of several characteristic joints in the abnormal posture image frame; The motion posture abnormal data unit is used to acquire and collect several characteristic abnormal joints of the abnormal posture image frame to obtain motion posture abnormal data.

10. The motion posture assessment system based on image processing according to claim 6, characterized in that: The posture adjustment module includes a posture adjustment unit; The posture adjustment unit is used to acquire several abnormal posture image frames in the motion video, acquire motion posture abnormality data of the several abnormal posture image frames, mark several characteristic abnormal joints in the several abnormal posture image frames and corresponding target reference posture image frames, respectively, and generate the motion posture adjustment data of the trainee, and display it to the trainee in the program.

Images