遮挡情况下的多目标跟踪算法*

在视频监控系统中，由于背景的复杂变化，运动目标经常会出现部分或全部被遮挡的情况。为了在遮挡条件下进行多目标跟踪，针对运动目标发生遮挡情况下的Mean Shift跟踪问题进行了研究，提出一种新的抗遮挡算法。利用卡尔曼滤波器来获得每帧Mean Shift算法的起始位置，再利用Mean Shift算法得到目标跟踪位置，通过目标遮挡判定机制和目标搜索机制来解决遮挡问题。实验表明，该算法较好地解决了运动目标的遮挡问题。

Tracking multiple objects in occlusions

In the video surveillance system, moving target often appears partial occlusion or complete occlusion due to the complexity of the background. To track multiple objects under the condition of occlusion, specific to the tracking of moving targets using Mean Shift algorithm in occlusion, this paper presented a new method with occlusion for object tracking. At first, found the starting position of Mean Shift with Kalman filter in every frame, and then utilized Mean Shift to track the target position. Used the mechanism of occlusion judgment and object searching to solve occlusion. Experimental results show that the algorithm is robust to solve occlusion problem.