Abstract: Aiming at the problems that the marks can not be matched and tracked in video motion parameter capture system, a 3D stereo multi-targets tracking algorithm is proposed, which locates the position of the marks by using 2D and 3D Kalman filters prediction. This algorithm also adopts 3D stereo matching method and the constrain relationship of epipolar line to complete matching and tracking of multi-targets. Experimental results demonstrate that the algorithm is effective.

Key words: multi-targets tracking, Kalman filter, 3D stereo tracking

摘要： 针对视频运动参数捕获系统中标记点无法被正确匹配及跟踪的问题，提出一种多目标三维立体跟踪算法。该算法通过二维和三维卡尔曼滤波预测来标记点的位置，并利用三维立体匹配方法以及外极限约束关系，解决了多标记点的匹配和跟踪问题。实验结果表明，该算法是有效的。

关键词: 多目标跟踪, 卡尔曼滤波, 三维立体跟踪

CLC Number: 

• N945.13