Sports Video Analysis And Personalized Customization

Recent years, with the great commercial value and enormous application potential, sports video analysis has attracting more and more researchers'attention, and has being a hot research topic in multimedia research region. Currently, the objectives and applications of sports video processing and analysis include:low-level feature analysis, mid-level semantic information extraction, high-level semantic event reasoning, highlight detection and extraction, video summarization, browsing, video retrieval, personalized customization, video content editing, enhancement and enrichment, etc.This thesis focuses on mid-level feature extraction (such as shot classification, replay detection), and multimodal integration method based highlight detection and extraction in sports video. The details are as follows:(1) In order to classify the video shot type efficiently, a decision tree based semantic shot classification is proposed. Firstly, several important shot types are prior defined. Then, color, texture and shape features are combined to perform view classification in the video sequences. And finally, the final shot type is voted by views contained in the shot.(2) In order to automatic detect the replay scene accurately, a replay-logo and shot context based approach is proposed. This method can not only accurate locate the replay boundaries, but also robust recognize replay scenes.(3) Based on shot type classification and replay scene detection, a multimodal integration method is employed to detect and extract the highlight of sports video. After the detection, the highlight can be annotated by the high-level semantic information, which is extracted from web-casting text and include play's name, player's team name, event type, etc. Based on those annotation results, a video database is established.(4) Based on video semantic analysis and annotation, some exploratory research on sport video personalized customization has being carried on, and a mobile based sport video customization system is designed and implemented, which can provide mobile users with interesting video clips according to their personalized preferences.