The application of edge feature in automatic sports genre classification

As a specific application of semantic video content analysis, automatic video classification has emerged as a very active area of research during the past few years. In terms of sports genre classification, commonly utilized features include color, motion, audio, and caption text. Although the edge feature is widely employed in other fields such as object detection, image enhancement and restoration, its potential value is underestimated, and it is seldom explored in automatic video content analysis. In this paper, we propose a sports video categorization method using edge feature. Our experiments show that our proposed method has achieved 97.1% accuracy on a set of 5 different popular sports video types. Moreover, we demonstrate the effect of video sequence length in accurate identification, and the advantages of edge feature over color information in sports genre classification