Abstract:

With the rapid growth of the video data’s amount, how to efficiently retrieve useful information has become very urgent. As the base of video indexing and searching, video annotation has great significance for its application prospect and research value. In the semantic detection, moving object detection and tracing is the basis. In the paper, adaptive Gaussian Mixture Model is used to background model; Camshift and Kalman filter are used to trace the players and ball. The implement of the algorithms is all based on Visual C++ and Visual c#2008. OpenCV and Aforge.net class base are also used. Experimental result shows that the method annotates well.

Key words: Sport Video, Kalman filter, Camshift algorithm, object detection and tracing