Statistical analyzing method of camera motion parameters for categorizing sports video

In this paper, we propose a statistical analysis method of camera motion parameters for categorizing broadcasted sports videos. We define two new features in the proposed method; "Camera motion extraction ratio" and "Camera motion transition". Camera motion parameters in the video sequence contain very significant information for categorization of broadcasted sports video, because in most of sports video, camera motions are closely related to the actions taken in the sports, which are mostly based on a certain rule depending on types of sports. Based on the characteristics, we design a sports video categorization algorithm for identifying 6 major different sports types. In our algorithm, the features automatically extracted from videos are analyzed in a statistical manner. The experimental results show a clear tendency of the applicability of the proposed method for sports genre identification.