DEFI-Net: Dual-Enhanced Feature Integration for Accurate Multi-Object Tracking in Sports Analytics

Introduction: Multi-object tracking (MOT) technology has significant applications in intelligent sports event analysis, enabling precise tracking and behavior recognition of athletes without human intervention. However, current MOT systems often face two major challenges in high-density interactions and fast-changing dynamic backgrounds. The first challenge is tracking discontinuity caused by motion blur and multiple occlusions in complex backgrounds, and the second is the difficulty in distinguishing targets due to the high similarity between target and background information. To address these issues, we propose a Dual-Enhanced Feature Integration Network (DEFI-Net), which combines background separation and motion prediction strategies to achieve comprehensive improvements in tracking stability and recognition accuracy. Methods: Firstly, to solve the issue of tracking discontinuity resulting from motion blur and multiple occlusions, we designed a Background Separation Adaptive Module (BSAM). This module leverages adaptive separation techniques to distinguish dynamic backgrounds from target areas, thus reducing background interference and ensuring continuous tracking in complex backgrounds. Secondly, to enhance recognition accuracy when target and background are highly similar, we introduced a Motion Prior Fusion Module (MPFM), which captures historical motion patterns and spatial position priors to accurately predict target positions, improving the model’s ability to differentiate similar targets and track them accurately. The innovation of DEFI-Net lies in its dual enhancement through background separation and motion prior integration, enabling robust tracking performance in high-density motion and complex backgrounds. Results: Experimental results show that DEFI-Net achieves significant performance improvements on public datasets such as LSP and SportsMOT, particularly excelling in scenarios with multiple occlusions and high background similarity. Conclusions: DEFI-Net technology significantly improves intelligent sports event analysis by addressing challenges like tracking discontinuity and target-background similarity, ensuring stable tracking and accurate recognition even in complex environments. Its practical application enhances real-time performance analysis, tactical decision-making, and injury prevention by enabling precise athlete tracking in high-density and dynamic settings.