Development And Practice Of Golf Assisted Training System Based On Video Analysis

In recent years,the development of computer vision technology has made significant progress,resulting in many platforms and systems that combine computer technology and sports assistance training,including intelligent systems that are combined with golf training and teaching.However,the existing golf assisted teaching intelligent system usually uses three-dimensional depth information,which will significantly increase the cost of the intelligent system.Of course,there are some golf-assisted teaching systems whose structure is too simple and lacks in digitization and intelligence.Therefore,this article takes intelligence and specialization as the goal of the system,and designs an efficient,professional,and cost-effective golf auxiliary training system In this thesis,the golf club slope is extracted on the basis of using a common monocular camera to collect golf motion video,in order to use the club slope information to match the professional coach swing video information.And at the same time,use OpenPose to capture two dimensional coordinate information,and process and analyze these data in order to apply it to the auxiliary training of golf swing.Using OpenPose for human body capture has the advantages of simple data processing,low equipment price,and no need to attach marker points on the human body.Specific technologies include the following three aspects:(1)An algorithm for golf club slope extraction based on edge information is proposed.First of all,on the basis of obtaining the club's center of mass and the club's connection,that the head of the golf club is more prominent than the shaft of golf club,,the original club's center of mass may be correcting the positioning point of the club has solved the problem that the center of mass of the club deviates from the shaft.Then,in view of the problem that the original club connecting area is too small,which leads to inaccurate extraction of subsequent club slopes,the paper uses the feature that the club is almost away from the human body,and re-screens the club area in order to select the appropriate club connection.region.Finally,using the club's edge information,the club slope is extracted.(2)A golf swing evaluation algorithm is proposed.First,considering the scalability of timing between different videos,in order to ensure that the template video and the user video have the same frame length before motion evaluation,the minimum distance video frame matching is performed on the template video and the user video.At the same time,in order to facilitate the matching between video frames,it is necessary to complete the joint point information by using the projected approximate point algorithm,and use the characteristics of the fixed human hand joints and fixed golf swing to complete the swing video.Segmentation.Then,in order to solve the problem that human joints have the same joint angle under different actions,the human limb angle is defined,and then the swing motion in the user video frame is evaluated(3)A golf assisted teaching system is designed based on the above research results.First,use OpenPose to extract two-dimensional coordinate information of human joint points from a golf swing video shot by a monocular camera.Then use the club system slope extraction algorithm to extract the slope of the club on the motion track,and match the video frames according to the club slope information,and finally use the human pose to evaluate the user's swing as a whole.The golf system designed in this paper has good practicability in real scenes.