A standard correction method for volleyball technical movements

Abstract

The present invention relates to a standardized correction method for volleyball technical movements. First, a batting video of a practitioner is recorded, key frames or pictures of the practitioner and standard movement videos are extracted, and the practitioner's movements and standard movements are identified using the OpenPose algorithm to identify key points of human body postures; then, the angle values ​​of the angles between specific joints of the human body are calculated, and the differences between the practitioner's movements and standard movements are compared, so that the practitioner's movements are corrected according to the analysis results. By using the method of the present invention, volleyball practitioners can not only prevent injuries caused by wrong movements, but also improve their volleyball technical level in a more scientific way, thereby laying a solid foundation for volleyball technology.

Description

Technical Field

[0001] The present invention relates to the technical field of motion analysis, and in particular to a method for standardizing and correcting volleyball technical movements. Background Art

[0002] In recent years, volleyball has become more and more popular among the masses. Volleyball is a sport that requires comprehensive skills and a high degree of skill. Each player must master various offensive and defensive techniques, so it is particularly important to have solid basic skills. For novices, it is very difficult to learn volleyball skills by just watching teaching videos, and it is also easy to form bad movement habits, making it more difficult to correct them in the future. The present invention can effectively standardize and correct the wrong movements of volleyball practitioners, and can propose improvement directions and basic movement essentials of the movement, which helps volleyball enthusiasts quickly master the basic movement postures of volleyball, and also avoids many practitioners from causing training injuries due to improper movements. Summary of the invention

[0003] The purpose of the present invention is to provide a standardized correction method for volleyball technical movements, which can utilize OpenPose to work well in complex spatial interferences such as contact, occlusion, and cutting, and utilize a bottom-up approach to achieve real-time and high precision, and realize the correction of volleyball technical movements by comparing the practitioner's movements with standard movements.

[0004] To achieve the above object, the technical solution of the present invention is: a method for standardizing and correcting volleyball technical movements, by comparing the movements of the practitioner with the standard movements, comprising the following steps:

[0005] Step S1, recording a volleyball player's action video, and extracting key action video frames or pictures of the player and standard action;

[0006] Step S2: Use OpenPose to identify the video frame, obtain the human body posture, and extract multiple key points of the body skeleton;

[0007] Step S3, calculating the joint angle values ​​of the practitioner's movements and the standard movements;

[0008] Step S4, comparing the angle values ​​of the practitioner's movements with those of the standard movements, and providing correction suggestions and tips for mastering technical movements.

[0009] In one embodiment of the present invention, in step S2, the OpenPose is the first open source real-time system for multi-person 2D posture detection, including hand, foot and facial key points, realizing real-time detection of multi-threaded key points; using the CPM deep neural network structure, combined with third-party dependency libraries including Caffe, CUDA, and CUDNN, it can accurately identify human skeleton points from ordinary images and determine the posture of the human body.

[0010] In one embodiment of the present invention, in step S2, the OpenPose is a bottom-up human posture estimation algorithm, and the specific steps include:

[0011] Step A1: Use the convolutional neural network CNN to analyze the input image, and initialize and fine-tune the first ten layers of the VGG-19 network to generate a set of features F that are input to the first layer;

[0012] Step A2: Feature F is processed through a continuous multi-stage network. The first part uses a convolutional network to predict the heat map S of the human body posture joints. t , and the other part uses a convolutional network to obtain the partial affinity domain L of all connected joints t ;

[0013] Step A3, starting from the second stage of the multi-stage network, the input of the stage t network includes three parts: S t-1 , L t -1 , F;

[0014] Step A4: In order to guide different branches of the network to perform different prediction tasks, different loss functions are used. The functions are as follows:

[0015]

[0016]

[0017]

[0018] in, is the true label value of the heat map, is the predicted value of the heat map at different stages; is the true label value of some affinity domains, is the predicted value of the partial affinity domain at different stages; W(P) is to prevent the data from being labeled incompletely, which may cause the loss function to have a particularly large value; J is the number of key points, and C is the number of limbs;

[0019] Step A5: Through multi-stage prediction, when the heat map and partial affinity domain are obtained from the final stage, the position information of all relevant nodes and the direction vectors of connected joints are obtained; then the Hungarian greedy algorithm is used to solve the optimization problem of the total affinity of the connection between the two types of key points, and then the posture of the whole body is assembled based on the connection of the same key points.

[0020] In one embodiment of the present invention, in step S2, a set of key points of a human body posture in a video frame is identified using a BODY_25 model.

[0021] In one embodiment of the present invention, the body key point recognition of the BODY_25 model includes the key points of 0-nose, 1-neck, 2-right shoulder, 3-right elbow, 4-right wrist, 5-left shoulder, 6-left elbow, 7-left wrist, 8-hip center, 9-right hip, 10-right knee, 11-right ankle, 12-left hip, 13-left knee, 14-left ankle, 15-right eye, 16-left eye, 17-right ear, 18-left ear, 19-inside of left foot, 20-outside of left foot, 21-left heel, 22-inside of right foot, 23-outside of right foot, and 24-right heel.

[0022] In one embodiment of the present invention, in step S3, based on the human key points obtained by OpenPose, the length between two key points can be obtained by calculating the Euclidean distance, and then the angle value of the angle between the joints of the human body movement can be calculated using the cosine theorem.

[0023] In one embodiment of the present invention, in step S4, since the OpenPose recognition accuracy is very accurate, the joint angle of the practitioner and the joint angle of the standard action are set to be allowed to differ by a predetermined value, and this value can be personalized according to the requirements of the trainee to achieve better training effects.

[0024] In one embodiment of the present invention, different actions require different joint angles to be compared; for the serving action, the right foot and the left foot are required to be approximately 30° apart, with the left foot in front and the right foot behind; for the blocking action, the entire arm is required to be straightened, and when hitting the ball, the arm is raised at approximately 90° to the body.

[0025] In one embodiment of the present invention, in step S4, when there is a large difference between the practitioner's action and the standard action, the difference is marked, and the technical essentials and training methods of the technical action are output.

[0026] Compared with the prior art, the present invention has the following beneficial effects: the present invention uses the OpenPose algorithm to compare the volleyball practitioner's movements with the standard movements, and then corrects the volleyball practitioner's incorrect movements. The present invention can effectively standardize and correct the volleyball practitioner's incorrect movements, and can propose improvement directions and basic movement essentials of the movement, which helps volleyball enthusiasts quickly master the basic volleyball movement postures, and also avoids many practitioners from causing training injuries due to improper movements. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is the OpenPose network architecture diagram;

[0028] Figure 2 is a system workflow diagram of an embodiment of the present invention;

[0029] Figure 3 It is a diagram for comparing volleyball assist movements. DETAILED DESCRIPTION

[0030] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings.

[0031] The present invention provides a method for correcting volleyball technical movements in a standardized manner, which corrects the wrong movements of the practitioner by comparing the practitioner's hitting movements with the standard movements, such as Figure 2 As shown, the detection method comprises the following steps:

[0032] 1. Record the action video of volleyball practitioners;

[0033] 2. Extract the video frames of the practitioner hitting the ball in the recorded video, and then extract the video frames of the same hitting action in the standard action video;

[0034] 3. Use OpenPose( Figure 1 The OpenPose network architecture diagram) algorithm is used to identify key points of the human body;

[0035] 4. Calculate different joint angles according to different technical movements;

[0036] 5. Compare the difference between the angle of the practitioner's hitting action and the standard action angle. If the angle difference exceeds the set threshold, it is determined that the practitioner's action is not standard, and corrective suggestions and the basic essentials of the technical action are output.

[0037] like Figure 3As shown in the figure, the left picture is a video frame of a volleyball standard training course of dribbling, and the right picture is a video frame of a trainer's incorrect dribbling action. As shown in the left picture, when dribbling, the angle between the thigh and the calf should be about 45°, the arm should be straight, and the angle between the upper arm and the forearm should be 180°; while in the right picture, the angle between the thigh and the calf of the trainer is 180°, and the angle between the upper arm and the forearm is also less than 180°, so it is judged as a non-standard action, and then a correction prompt "Arms are not straightened, legs are not bent" is given, and then the training essentials of the dribbling action are given: half squat in the preparation posture, and relax the whole body. Straighten both arms and clamp them, press the wrists down, and bend the knees slightly. When hitting the ball, the arms are not bent. The most important point is the softness of the arms. The arms that are clamped straight should also be softly raised at the moment when the ball falls on your wrists, and then find a good landing point for the ball to hit back.

[0038] The above are preferred embodiments of the present invention. Any changes made according to the technical solution of the present invention, as long as the resulting functions do not exceed the scope of the technical solution of the present invention, belong to the protection scope of the present invention.

Claims

1. A method for correcting volleyball technical movements by comparing the movements of the practitioner with the standard movements, characterized in that: The following steps are involved: Step S1, recording a volleyball player's action video, and extracting key action video frames or pictures of the player and standard action; Step S2: Use OpenPose to identify the video frame, obtain the human body posture, and extract multiple key points of the body skeleton; Step S3, calculating the joint angle values ​​of the practitioner's movements and the standard movements; Step S4, comparing the angle values ​​of the practitioner's movements with those of the standard movements, and providing correction suggestions and tips for mastering technical movements; In step S2, the OpenPose is a bottom-up human posture estimation algorithm, and the specific steps include: Step A1: Use the convolutional neural network CNN to analyze the input image, and initialize and fine-tune the first ten layers of the VGG-19 network to generate a set of features F that are input to the first layer; Step A2: Feature F is processed through a continuous multi-stage network. The first part uses a convolutional network to predict the heat map S of the human body posture joints. t , and the other part uses a convolutional network to obtain the partial affinity domain L of all connected joints t ; Step A3, starting from the second stage of the multi-stage network, the input of the stage t network includes three parts: S t-1 , L t-1 , F; Step A4: In order to guide different branches of the network to perform different prediction tasks, different loss functions are used. The functions are as follows: in, is the true label value of the heat map, is the predicted value of the heat map at different stages; is the true label value of some affinity domains, is the predicted value of partial affinity domain at different stages; J is the number of key points, and C is the number of limbs; Step A5: After the heat map and partial affinity domain are obtained from the final stage of the multi-stage prediction, the position information of all relevant nodes and the direction vectors of the connected joints are obtained; the Hungarian greedy algorithm is then used to solve the optimization problem of the total affinity of the connection between the two types of key points, and then the posture of the whole body is assembled according to the connection of the same key points; In step S4, the joint angles of the practitioner and the joint angles of the standard action are allowed to differ by a predetermined value, and this value is personalized according to the requirements of the trainee; the joint angles that need to be compared for different actions are different; for the serving action, the right foot and the left foot are required to be about 30 degrees, with the left foot in front and the right foot behind; for the pad action, the whole arm is required to be straight, and the arm is raised at about 90 degrees to the body when hitting the ball; in step S4, when the practitioner's action is significantly different from the standard action, the difference is marked, and the technical essentials and training methods of the technical action are output.

2. A method for correcting volleyball technical movements according to claim 1, characterized in that: In step S2, the OpenPose is the first open source real-time system for multi-person 2D posture detection, including hand, foot and facial key points, to achieve real-time detection of multi-threaded key points; it uses a CPM deep neural network structure, combined with third-party dependency libraries including Caffe, CUDA, and CUDNN, to determine the posture of the human body.

3. A method for correcting volleyball technical movements according to claim 1, characterized in that: In step S2, the key point set of the human body posture in the video frame is identified using the BODY_25 model.

4. A method for correcting volleyball technical movements according to claim 3, characterized in that: The body key point recognition of the BODY_25 model includes the key points of 0-nose, 1-neck, 2-right shoulder, 3-right elbow, 4-right wrist, 5-left shoulder, 6-left elbow, 7-left wrist, 8-hip center, 9-right hip, 10-right knee, 11-right ankle, 12-left hip, 13-left knee, 14-left ankle, 15-right eye, 16-left eye, 17-right ear, 18-left ear, 19-left foot inside, 20-left foot outside, 21-left heel, 22-right foot inside, 23-right foot outside, and 24-right heel.

5. A method for correcting volleyball technical movements according to claim 1, characterized in that: In step S3, based on the human key points obtained by OpenPose, the length between two key points is obtained by calculating the Euclidean distance, and then the angle value between the angles of the human body movement joints is calculated using the cosine theorem.

Images