篮球视频中球员的分割与跟踪算法

分割与跟踪一直是视频图像处理的一个热门问题。针对篮球视频中激烈运动的球员队属和跟踪问题，实现了一个实时的球员分割与跟踪算法。首先，不同于一般的分割方法，采用基于两个级别（像素级和帧级）实时分割和背景更新的背景减除方法，实现了对球员的分割；接着，通过将球员与模板进行马式距离比较来辨识球员队属；然后，使用基于核的目标跟踪方法，实现了在激烈比赛中对球员的准确跟踪。实验结果表明，提出的算法能够很好地辨别各个球员的队属问题并在不完全遮挡的情况下跟踪多个球员，而且在球员形态变化较大时也能稳定地跟踪。

Tracking of players from basketball video sequence

Segmenting and tracking are a challenging problem in video processing.To solve the problem of identification and tracking of the players in drastic basketball game，this paper presents a real-time method to implement segmentation and tracking of players from a basketball video.Different from other common segment method，players are segmented by the use of the background subtraction based on two level（pixel level and frame level）；Mahalanobis distances between a player and a model are used to identify the team that the player belongs to；This kernel-based object tracking method has been used to track multiple players in a stinging basketball game.Experiment results show that this real-time method deals well with identification of the players，and can track multiple players in partial occlusion，and can track robustly when the players’ shape changes greatly.