Research on athlete target tracking based on simple online and realtime tracking with a deep association metric algorithm

As the computer technology rapidly develops, the application of target tracking technology in the field of sports is getting more and more attention. In order to better analyze the athletes’ movement state, research has been conducted on athlete target tracking technology based on deep correlation metric tracking algorithm. Firstly, a suitable dataset is selected and the video data is preprocessed with image enhancement and denoising. Then the attention mechanism is introduced, motion prediction matching is performed using Kalman filtering, and the Hungarian algorithm is applied to initialize the cost matrix and calculate the matching cost. Target detection is applied in each module of the algorithm for improvement and the loss function is redefined. High tracking accuracy and trajectory continuity are realized by experimental testing, and the algorithm achieves 97% trackin g accuracy in the tracking phase, while the trajectory continuity is raised by 14.6% and 6.5% respectively in contrast to other algorithms, which is a better performance state. The designed athlete target tracking method based on the deep correlation metric algorithm can get more continuous trajectory data while tracking accurately, which is helpful for the intelligent development of sports programs.