Automatic tracking method for sports video analysis

This paper addr'esses t.he implemenl.ation oj an advanced visual-based t.ennis sports video analysis system. Firstly, a whit.e pi1;el-Jraction based met.hod is used to detect. the significant Jrames that include the t.ennis court. In addition, we employ the I.emporal correlation between two consecutive Jrames 1.0 track the court locat.ion within a local search ar'Ca" Furtherm.01T, 'We p1'Opose a player segmenl.at.ion and I.racking algor'ithm thai. sepam./.ely builds backgmund models J01' the playing Jield and the ar'm s1Lrrmmding t.he .field according t.o their' diffeTent. color's. This enables 1.0 pmvide it. as a pr'CIlict.ion oj the CUTTenl. frame JOT mOTe efficiency. ResuUs on real tenn'is video data ar'e presented, demonsl.rating t.he validity and perI01-m.ance of the appmach