Tracking Players in Low-Resolution Videos of Soccer Games

This thesis concerns the topic of tracking soccer players. The problem is to select the best from relevant methods for tracking a single soccer player in a low-resolution video sequence captured by a single static camera. Achieving this is the basis for developing a system for tracking soccer players in low-resolution videos with multi-camera approach, which is the project assigned by the Software Competence Center Hagenberg (SCCH, Austria) in collaboration with the Center for Machine Perception (CMP, Prague). The kernel-based object tracking method with background-weighted histogram and motion prediction was implemented. It outperformed other related state-of-the-art methods with about 90% tracking success rate on five soccer sequences. It achieves a real-time processing speed of 250 frames per second.