一种基于背景加权直方图的Mean-Shift跟踪算法

A Mean-Shift Tracking Algorithm Based on Background Weighted Histogram

经典的Mean-Shift算法以颜色直方图为特征进行跟踪,但在视频目标跟踪中没有有效地利用跟踪目标的空间信息,因而当目标快速移动时,目标极易丢失。针对这一问题,提出了一种基于距离目标形心加权直方图的Mean-Shift跟踪算法。距离目标形心加权直方图可有效利用目标在图像中的位置信息,从而能实现复杂背景下的目标跟踪。并分别用改进的基于距离目标形心加权直方图的Mean-Shift跟踪算法和经典的Mean-Shift跟踪算法,对快速运动的目标进行跟踪实验。结果表明,改进的基于距离目标形心加权直方图的Mean-Shift跟踪算法,具有较强的鲁棒性和实时性,能有效实现复杂场景下的目标实时跟踪。

The classic Mean-Shift tracking algorithm, characterized by the color histogram, does not make full use of the position information of the target, therefore it is easy to lose the target as the target fast moving. Proposes a Mean-Shift tracking algorithm based on weighted histogram of the distance to target center.The weighted histogram effectively uses the target spatial information in the image to achieve the target tracking in complex background. Applies respectively the improved Mean-Shift tracking algorithm and classic Mean-Shift tracking algorithm for fast moving target tracking tests. The result shows that the improved algorithm has good robustness and real-time performance and effectively achieves the target tracking in complex scenes.