The Research On Golf Swing Action Comparison Based On Video Human Body Pose Estimation

In recent years,with the improvement of people's living standards,people are paying more and more attention to sports exercise.The traditional sports training mainly relies on the coach's trainingmethod.However,because the coach resource is scarce and the training cost of one-to-one or one-to-many coaches is high,it is mainly based on the motion capture system or carrying sensors.The collection of motion parameters not only causes a certain degree of inconvenience to the athletes,but also hinders the development of intelligent sports.Intelligent motion based on computer vision can solve these problems.For this reason,based on the golf swing video,a video-based human pose estimation algorithm and a golf swing gesture comparison analysis algorithm based on human body pose estimation for golf swing video are proposed.Aiming at the shortcomings of OpenPose,a video-based human pose estimation algorithm is proposed.The method firstly uses the OpenPose algorithm to perform static image human body pose estimation,then establish human tracking based on the inter-frame attitude distance metric and performs super pixel segmentation on the image to determine the super pixel of the bone joint point,The intersection area of the super pixel and the joint-centered box is taken as the minimum granularity,and then the superior joint point is searched by the forward and backward directions,and the candidate joint point set and the reference joint point are established based on the optical flow and the human motion continuity.Finally,optimal candidate key points are generated for the key points with lower confidence in each frame image,and the optimal global human body pose is generated for each frame image by reorganizing these body parts.The video-based human pose estimation method presented in this chapter has shown good results on the PoseTrack database.Compared with OpenPose,it has improved in each bone joint point,which improves the OpenPose error detection,missed detection and false detection.In order to analyze the swing action in the golf swing video,an intelligent comparison analysis algorithm of golf swing video is proposed,and the video human body pose estimation is used for the intelligent comparison analysis of golf swing video.First,the golf swing video is used to estimate the human body posture of the video,extract the information of the human skeleton joint point of the athlete,and perform corresponding pre-processing.then extract the trajectory characteristics of the elbow in the vertical direction according to the skeleton joint point data of the athlete,Then,the modified DDTW algorithm is used to segment the golf swing action sequence.Then,by using the relationship between the trajectory of the elbow in the vertical direction and the key action frame,a part of the golf swing key action frame is extracted,and then some artificial labor is extracted.The characteristics of the design are used to extract the remaining key motion frames,and finally the dynamic features and static features are extracted from the two sides of the positive side,and combined with the comparative analysis rules formulated by professional golf coaches for quantitative analysis.This chapter proposes a DDTW algorithm that can be used for behavioral sequence segmentation and a method for extracting key motion frames of golf swing based on motion trajectory.It can achieve good segmentation and key action frame extraction effects in the golf swing motion video database,and achieve good results in accuracy,finally it verifies its practicability and applicability.