遮挡情况下基于特征相关匹配的目标跟踪算法

特征相关匹配是重要的运动目标跟踪方法.目标特征有灰度特征和边缘特征两大类,在遮挡情况下,采用哪种特征进行匹配,要根据目标本身属性来确定.本文先对目标灰度性质做出判断,然后根据灰度单一或是丰富来合理选择边缘相关匹配或者是基于多子块的灰度相关匹配来解决遮挡情况下的刚性目标跟踪问题.其中边缘匹配算法是通过当前边缘与实时更新模板的最优匹配来确定目标的运动位移量.基于多子块的灰度相关匹配算法通过目标的各个具有较明显特征的子块准确判定遮挡区域,利用剩余的未被遮挡的子块参与灰度相关匹配继续跟踪目标.实验结果表明,这种算法是十分有效的.

Target Tracking Algorithm under Occlusion Based on Feature Correlation Matching

Feature correlation matching is a very important method for target tracking. Intensity and edge are two main kinds of target features. Under occlusion, the tracking feature should be chosen depending on the characteristics of the target. This paper estimates the intensity characteristic of target firstly, then choose the edge matching or intensity matching algorithm based on the narrow or wide intensity ranges of the target. The edge matching method determines the optimal displacement vector between two successive frames by matching of the target edge template and the current unoccluded edge, and the template is updated in real time. The intensity correlation matching based on multi-block method can estimate the occlusion region accurately through the blocks with distinct feature, and track the target by the remaining unoccluded blocks. Experimental results show the good performances of the proposed algorithm.