Visual analysis of sports multi-target trajectory based on hybrid motion tracking algorithm

The complexity of multi-target motion tracking in sports poses significant challenges due to frequent occlusion, rapid movement changes, and high interaction dynamics. To address these issues, this study proposes a hybrid motion tracking algorithm that integrates optical flow-guided particle filtering (OF-PF), a lightweight graph attention network (LGAT), and a biomechanical graph propagation network (BGPN) to enhance trajectory accuracy and tactical reasoning. The proposed framework was evaluated on the 1st and Future - Player Contact Detection dataset, achieving 92.4% tracking accuracy and reducing the ID switch rate to 2.1%, outperforming conventional methods. The results demonstrate the model's effectiveness in improving athlete movement prediction, tactical pattern recognition, and contact event reasoning. This research provides a novel visual analysis approach for automated sports analytics, enabling real-time tracking and decision support in competitive environments.