融合多线索的目标跟踪  

Object tracking by combining multi-cues

作　　者：朱兵[1] 李金宗[1] 魏祥泉[1] 

机构地区：[1]哈尔滨工业大学340信箱,黑龙江哈尔滨150001

出　　处：《计算机应用》2006年第3期601-604,共4页journal of Computer Applications

摘　　要：提出把目标的表观信息和目标的运动信息融合起来跟踪的一种方法,使用直方图来描述目标的表观信息,用背景配准来检测目标的运动变化,把通过这两种信息分别获得的定位融合起来得到目标定位,然后采用扩展卡尔曼滤波,实现有效的目标跟踪。该方法克服了在目标较小时,使用表观信息跟踪不稳定的缺点。实验结果也证明了在目标被部分遮挡,目标在像平面由小到大过程都能够稳定跟踪。A tracking algorithm combined appearance with motion of object was addressed. Histogram was used to describe the appearance of the object and background registration was implemented to detect the motion of object. The location of object was gained by fusing the two positions which were detected by the appearance and the motion of object respectively. Extended Kalman filter was used to track the object. With the help of this combination, the unstable performance of tracking only with the appearance when object was small was removed. Experiments show the stabilization of the tracking algorithm when target is occluded partly or target changes from small to big in the image.

关 键 词：直方图 图像配准 运动检测 跟踪 扩展卡尔曼滤波 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]