For a movement target segmentation complete video result, it can reduce movement target search quantity, and is of great help to effective promoting tracking speed and accuracy. The purpose of video segmentation is to extract movement target out from background in video sequences and used for implementing prospect and background segmentation. The paper researches on background deduction before tennis video human body movement tracking and contour extracting, in the hope of exploring more scientific background deduction algorithm to provide good platform for sports video analysis. In the paper, it firstly introduces Mean Shift algorithm and particle filter method algorithm. It states to algorithm that combines with the two algorithms, and provides LPP background estimation algorithm and particle filter prediction algorithm principle. Use the paper algorithms to compare 3 algorithm, and shows algorithm superiority.