Research And Realization Of Analysis Method Assisted By Semantic Shot For Soccer Video

With the development of video storage technique, speedy increase of users* demand and advancement of network, automatically detecting important semantic events of the soccer video and further summarization of the video to help indexing, browsing and presenting it has become increasingly important.But, there are lots of problems in soccer video analysis because of the complicacy of object motion and the loose temporal structure. Most of researches about soccer video analysis are in the phase of exploring methods for analyzing basic characters, lack of available methods for systematic analysis. This paper is mainly concerned with the systematic method for detecting and analyzing shot events of soccer video, provides a entirely novel and more generic method for investigating highlights. The major domains of the research are semantic shots model, semantic shots detection, detection methods about the event based on semantic shots and analyzing motion focus. The major content and contribution of the dissertation include as following:1. Studies the structure and production technique of soccer video, then puts forward and realizes a model about high-level semantic shot, and builds an effective relationship between low-level physical features and high-level semantic contents.2. It discusses some methods of shot detection. Simplifies and improves a content-based scene change detection and classification technique using background tracking and develop a valid method for shot detection.3. Based on the classification of semantic shot, after analyzing the sequence in time among various types of shot, building scene transfer chart, it gives and achieves a novel technique for shot event detection.4. It works on and realizes a way to analyze attack route based on analyzing motion focus, and resolves the problem on continuous ball detection and tracking.5. By integrating these above-mentioned researches, this paper forms a systematic technique. Linking with the annotation of semantics and individual retrieval function, we have developed a system of analyzing and managing highlights of soccer video. By experimenting these systems, we have gained satisfied effects.