The Design And Implementation Of A Sports Video Analysis System Based On The Target Tracking

The competition on sports arena of each country largely depends on the science andtechnology that is behind it. Compared with western countries, China relatively lagsbehind in sports technology. There is few scientific training adding tools for the coach toinstruct the athletes. With target tracking, the video analysis system of competitivesports can analyze the digital video in the sports events and trainings locate and track thetarget. Since the computer vision is more retentive and accurate, it will greatly improve thescientificity and efficiency of the sports training with extensive application.This paper mainly studies the basic theories and critical techniques. It gives a briefintroduction of methods of the feature extraction, target representation and localization inthe video tracking. Bayesian tracking method based on IMCMC is introduced in details,which decomposes the tracker into observation and motion model. The observation modelwas decomposed into several basic observation models built by a group of sparse principalcomponent analysis（SPCA）in the feature template. Each basic observation model coversthe special appearance of the target. The motion model is also the result of combination ofseveral basic ones with different types of motion models. Through the frame of interactiveMarkov Chain Monte Carlo（IMCMC）, it finally integrates the basic observation modelsand motion models into a compound tracker. Each basic tracker is in charge of specificvariation of the target and communicates with other trackers to realize accurate tracking.With the help of Bayesian tracking method based on IMCMC, I design a videoanalysis system of competitive sports by using the techniques of MFC、OpenCV、Matlaband etc. It mainly has five functions, namely, the contrast video synchronization,background synthesis of automatic camera, target tracking, output of motion track, andparameter setting, verifying the practicability of algorithm.