基于改进颜色直方图映射的目标跟踪算法

Object Tracking Using an Improved Color Histogram Back - projection

关于视觉跟踪技术问题与电目标背景颜色相关,针对于运动目标易受到背景颜色的干扰,提出了一种基于改进直方图映射的目标跟踪算法,为了抑止模型中混有背景颜色的干扰,并能准确定位目标。依据初始的目标模型(前景),以及跟踪过程中搜索范围(背景)的颜色直方图按照一定的比例关系建立起目标概率灰度值索引表,将索引表映射到跟踪搜索窗口中,采用meanshift算法在生成的灰度图中快速定位目标位置。实验结果表明,改进的颜色直方图映射算法能够从根本上抑止模型的背景干扰,并且meanshift均值漂移算法能够准确的定位目标。

Moving targets are liable to be interfere with by the background color.An improved histogram backprojection is proposed in this paper.The purpose of the algorithm is to suppress background interference fundamentally.According to the initial model（foreground） and the color histogram of search area（background） during the tracking course,a gray value index table of objectives probability was established,which was mapped to track the search window.Mean-shift algorithm was used to locate the target in the generated grayscale image quickly,which can find the local maxima to locate the object near the neighborhood of object center in the previous frame.Experimental results have shown that the algorithm can suppress background interference fundamentally and track the moving object precisely in real-time.