Answers are grounded in this paper and cite their sources.
Demonstrates enhanced tracking accuracy and prediction in high-speed ball sports, indicating a robust solution for event analysis.
In high-speed and competitive sports such as football and basketball, real-time tracking and trajectory prediction of players and the ball are core technologies for intelligent event analysis, assistance in judging penalties, and immersive live broadcasts. Existing methods suffer from unstable tracking in complex scenarios like strong occlusion and high-speed directional changes, as well as separation between prediction and tracking, and insufficient adaptability. To address these challenges, this paper proposes a tracking and prediction coupling algorithm. Based on deep learning and temporal modeling, this algorithm constructs a three-layer architecture of feature fusion, coupling interaction, and adaptive adjustment, enhancing target identification through multi-dimensional feature fusion, establishing a bidirectional feedback mechanism to suppress error accumulation, and dynamically adapting to scene changes. Experiments were conducted using the SoccerNet, BasketballNet public datasets, and a self-built high-speed confrontation dataset. The results show that the proposed algorithm achieves a tracking accuracy of 95.2%, with a trajectory prediction error of only 0.87 ** pixels, and a processing frame rate of over 35 frames per second. In scenarios with severe occlusion and high-speed movement, the tracking accuracy increases by 8.3%–12.7%, and the prediction error decreases by 31.2%–42.6%. The performance is stable across different confrontation intensities, and the overall performance is significantly superior to mainstream and latest sports-specific comparison algorithms. This work presents a promising integrated tracking-and-prediction framework for high-speed adversarial ball sports video analysis, providing an efficient and robust solution for intelligent event analysis, applicable to tactical analysis, intelligent training, and enhancement of real-time live broadcasts.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: