Development of a multi-level feature fusion model for basketball player trajectory tracking

• DOI: 10.1016/j.sasc.2024.200119
• Publication Date: 2024-07-04T00:10:35Z
• Authors: Tao Wang
• Keywords: Basketball; Tracking (education); Feature (linguistics)

To solve the problems of low matching degree, long tracking time, and accuracy multi-target in process athlete motion trajectory using deep learning technology, a new model was proposed this study. The study first optimized current object detection algorithm basketball, utilized hybrid attention mechanism to extract features, improved non-maximum suppression strategy. Then, branch network introduced improve residual identity recognition proposed. Finally, designed by combining model. research results indicated that experiment, time always below 0.4 s, its average reached up 0.63. In testing, final built had 0.98, overlap rate as 0.02. This has following two contributions. Firstly, is proposed, which improves efficiency optimizing introducing network. Second, excellent performance both track tracking, can not only provide solution for athletes' but also significantly effect tracking.