An effective trajectory-based algorithm for ball detection and tracking with application to the analysis of broadcast sports video

Abstract

This thesis is on sports video analysis and enhancement. It addresses three closely-related problems. It first addresses the ball detection and tracking problem in broadcast sports video. It proposes an effective trajectory-based algorithm for locating the ball in a broadcast sports video, which can obtain the accurate results for broadcast soccer/tennis video. Then, it addresses two applications of ball locations: ball-related event detection and reconstruction and enrichment of broadcast soccer video (BSV). For the first application, it proposes a trajectory-based event detection approach, which improves the event detection performance because the events have a close correlation with the ball location. More importantly, this approach can detect some events that may not be detected if we just use low-level features. For the second application, it proposes a reconstruction and enrichment system for BSV. In addition, this thesis proposes a robust ellipse Hough transform and applies it to detect ellipses in BSV.