Research and Implementation of Wushu Video Retrieval Based on Semantic

This paper put forward a semantic-based video analytic approach and it´s two algorithms in the Wushu video retrieval. Firstly, it can be taken a edge detection to the special-temporal slices by using the edge detection operator based on bilateral filtering, and then extracting and training the classes used the key frame semantic based on SVM. The system which is completed by the two algorithms implement the segmentation and retrieval of the Wushu video. The experiment show that it is better to retrieval according to the characters of the Wushu video by using this system.