一种有效的随动跟踪方法

An Effective Method for Follow-up Tracking

针对视频监控系统,提出了一种离散化像机运动的随动跟踪方法,解决了像机运动时对目标检测带来的困难,并弥补了由于单个摄像机拍摄范围的局限性。文中针对像机监控场景,离散化像机转动,建立索引表;通过多帧差分和自适应背景方法提取运动对象,并结合K alm an滤波器和多帧预测思想实现了对特定目标的鲁棒跟踪;综合利用目标的信息、索引表和云台实现了对特定目标的随动跟踪。实验结果表明,该算法对于复杂场景中运动目标的检测和跟踪具有较好的鲁棒性和实时性;对特定目标的随动跟踪具有良好的实时性和稳健性。

In the full paper, we explain in detail our effective method around what we consider as the three crucial problems in follow-up tracking; in this abstract, we just sketch an outline of our explanation. The first crucial problem is the establishment of two index tables through discretization of motion of camera： （1） rotation index table for pan motion （or rotation in the horizontal plane） and for tilt motion （or rotation in the vertical plane）; （2） background index table. The second crucial problem is, with the camera fixed in position, the real-time detection and tracking of moving targets; the tracking is done by a specially designed Kalman filter; particular attention is paid to solving the occlusion problem. The third crucial problem is the scheduling of the pan and tilt motions of camera. Experimental results show that our method is robust and effective in follow-up tracking of interested target with complex background, and can extract moving targets real-time.