Research on key event detection and tactical analysis in sports videos via spatio temporal graph network

To address the challenges in key event detection and tactical analysis for sports videos, this paper proposes a unified analysis framework integrating a Spatio Temporal Graph Network (STGN). The method first obtains positions and trajectories of players and the ball through multi object tracking, then dynamically constructs a sequence of heterogeneous graphs including player player spatial edges and player ball possession edges. A spatio temporal graph encoder is then used, stacking heterogeneous graph convolutions and gated recurrent units. Finally, a dual task prediction head is designed to jointly optimize event classification, tactical pattern recognition, and player role classification. Experiments show that the proposed method achieves a mean average precision of 79.5% for key event detection, a macro average F1 score of 80.7% for tactical pattern recognition, a silhouette coefficient of 0.62 for player role clustering, and an inference speed of 52.8 FPS. All metrics significantly outperform competing methods such as SlowFast, VideoMAE, and ST GCN. The proposed spatio temporal graph network can effectively model dynamic relationships among multiple entities, providing an accurate, efficient, and structurally interpretable technical solution for intelligent sports video analysis.