Jump-Aware: Player Position Rectification and Identification in Dynamic Sports Using Jump Event Spotting

Yin May Oo, Ankhzaya Jamsrandorj, Vanyi Chao, Hoang Quoc Nguyen, Yewon Hwang, Kyung-Ryoul Mun, Jinwook Kim

Abstract: Accurate player positioning and identification in broadcast sports videos are essential for sports analytics. However, jump-intensive sports such as basketball and volleyball pose significant challenges due to positional distortions caused by airborne motion and occlusions. To address these issues, we introduce Jump-Aware Position Rectification (JPR), a framework that integrates Jump Event Spotting (JES) and jersey-based player identification to improve spatial consistency and identity tracking. Our method first detects and validates jump events, then rectifies player positions in a top-view pitch coordinate system, reducing motion artifacts caused by temporary elevation changes in 2D image space. Additionally, jersey-based identification enhances identity tracking by leveraging jersey numbers, even under occlusions. To support our research, we present SportsJumpMotion, a dataset featuring frame-accurate jump annotations and jersey-based player identities for basketball and volleyball. Experimental results demonstrate that our JES model achieves a mean Average Precision (mAP) of 95.33, outperforming baseline models in jump event spotting. Furthermore, fine-tuning on sport-specific datasets significantly improves jersey-based identification, addressing variations in jersey visibility and motion patterns across sports. Our dataset and framework provide a comprehensive benchmark for advancing player tracking in dynamic sports scenarios. Our SportsJumpMotion dataset is publicly available at https://github.com/yinmayoo185/SportsJumpMotion.