An Identification Algorithm for Rigid Athlete Numbers in Sports Video

Recently, sports video analysis for retrieval purpose has become a hot research area. As important clues to sports video content analysis, detection and recognition of the number of athletes is a direct method to identify athletes. In this paper, we propose a method of detect rigid number of athletes. The number of athletes, as an example, we explored methods of detection and identification of the rigid deformed characters. First, we adopted a method that combines edge features and color features to locate the athlete number cloth. Then we used perspective transformation for correction of deformed area to identify the rigid deformation characters. Finally, for the candidate regional characters after adjustment, we employed the method of character template matching for identification. Experimental results on typical sports images show the potential of our algorithm.