Probabilistic Integration of Tracking and Recognition of Soccer Players

摘要: This paper proposes a method for integrating player trajectories tracked in wide-angle images and identities by face back-number recognition from motion-controlled camera. In order to recover tracking failures efficiently, the camera scans follows players who are judged likely undergo heavy occlusions several seconds future. The candidates of each trajectory probabilistically modeled updated at every identification. degradation due passage time also modeled. Experiments showed system's feasibility automatic real-time formation estimation which will be applied metadata production with semantic dynamic information on sports scenes.