Video High-level Structural Analysis And Mining

With the development of the compress and transport technique, video is widely used. A great lot of digital video databases appear, which take rich and colorful video data to the people. Recent applications of video focuses on video on demand, content based retrieval system and video summary. These applications solve the problem of finding and retrieval video recourse, but it can't find the valuable "knowledge" contained in video data, so we must research the new analysis method for video data. This method is named video mining.This paper has researched definition, prototype and technique of video data mining and the high structure mining of video. The mostly work is as follow.1. The paper bas discussed the concept and prototype of video data mining. As to the peculiarity of video data, the paper has analyzed the technique in common use of mining, such as classification, clustering and association. The foreground of video data mining is discarded, too.2. To supply the video data mining, the paper has analyzed the multi-character of video data. Firstly, segment the video into shots, then analyze the character, like color, texture, shape, semantic information and movement.3. The paper has researched high structure of video. Use the improved clustering method to find knowledge of video structure, and then use the knowledge to conformation the hiberarchy of video.4. The paper has research two problems associated with the video structure. The one is frequency of shots change. The other is the narrative structure of video program.Finally, I do experiments to validate the video structure mining method that this paper brings forward, and the result is satisfying.