Automatic Classification System of Wushu Video Based on SVM

This paper presents a semantic-based video analysis method and two steps of their classification in the wushu video: Firstly, used of the image frame difference to construct on the basis of background image, and used of threshold settings to determine the body information and background information, This can achieve the elimination of background and exercise the purpose of extracting the human body. The system which is completed by the two algorithms implement the classification of the wushu video. The experiment show that it is better to classify according to the characters of the wushu video by using this system.