The Detection Of Net-Approach Event In Tennis Video

Tennis is a very competitive spectator sport game. It becomes more and more popular among people in recent years. Therefore how to detect the specific events in its playing video effectively has becomes a widely popular research topic. Particularly, in tennis the Net-approach event is a perfect play that draws intensive spectators’ attention in which the paper taken as the research topic with primary objective of how to automatically detect the Net-approach motion.First of all, extract the moving object in tennis video. According to the characteristics of live tennis video, we proposed an improved algorithm that based on Gaussian Single Model, and detect the mobile target by using the background subtraction method. The experiment showed that the improved method can effectively decrease the detection errors caused by the camera motion, and reduce the time of rebuild Gaussian Single Model when the shot changed.Secondly, locate the position of athletes. After we identify the moving targets, in order to detect the occurrence of the Net-approach event precisely, we used the feature matching method to locate the athletes, then we contrast the advantages and disadvantages of SIFT and centroid feature. After the comparison, we choose the centroid tracking method that based on contour extraction, and realized the tracking of athletes.Thirdly, smooth and correction the trajectory. The trajectory generated by centroid is obvious jitter. It has a great error compared with the real motion trajectory of the athletes. So we used the Kalman Filter to eliminate the error. And it’s proven to be effective to the rectify detection of Net-approach event afterward.Finally, judge the occurrence of Net-approach event. By applying the relative change of the athletes position and the domain knowledge in tennis, this paper determines the start and end of the Net-approach event as well as storage the video frame. The results show that the effect of our method is quite productive.