Helmet Allocation And Tracking Methods For Football Players Based On YOLOv7 And Deep Sort

In recent years,American football has gradually come into the public eye.As a sport with frequent collisions,players wearing similar clothes are often difficult to recognize in the game.Therefore,in order to improve the auxiliary spectator experience and facilitate the health monitoring of collision players,it is of great significance to study the helmet tracking and numbering assignment of football players.In the process of football player helmet tracking and numbering allocation,this paper further optimized and improved the football player helmet allocation and tracking method based on YOLOv7 and Deep Sort algorithm,aiming at the problems such as obvious noise in some video frames,small proportion of player helmets,different frame rates of labeling in different dimensions,large number of camera positions and lack of labeling.The main work is as follows:(1)For the helmet detection and prematching method of sports players,the modified YOLOv7 target detection algorithm based on weighted boundary box fusion is used to integrate the prediction results of multiple detection models,improve the accuracy of prediction boundary box and reduce the missing detection of small target players;The rotation matching strategy is used to remove the relative coordinate offset of video frame and sensor frame.Adaptive frame clipping strategy is used to eliminate the interference of the helmet of the sideline players.(2)In terms of helmet tracking and number allocation,the base line and sideline perspective prediction strategy is used to solve the difficult problem of machine position prediction in the case of lack of tags.The dual axis matching strategy of x axis and y axis is used to further improve the accuracy of player number matching.The Deep Sort tracking method with enhanced feature extraction was used to improve the differentiation of players with different numbers,reduce the false detection rate of player tracking,and improve the problem of frequent switching of player identification in collision scenes.In this paper,based on YOLOv7 and Deep Sort’s method of helmet allocation and tracking for football players,relevant comparative ablation experiments were conducted for the above improvement points.The experimental results show that the improved method performs well in helmet tracking and matching for football players.