基于均值漂移的自适应跟踪算法

摘要

The tracking window of classic mean-shift tracking algorithm is fixed in the tracking process, so that it cannot accurately track the target whose tracking size changes frequently. An adaptive target tracking algorithm based on mean-shift was proposed in this paper. Initially the changes of the target are determined by using the edge weighted probability density. When the target was becoming bigger, the centroid positioning and the feature points affine model were used to update the tracking window. When the target was becoming smaller or unchanged, the Bhattacharyya coefficient was used to determine the reduction ratio of the target. Simulation results showed that the proposed algorithm could effectively track the target whose scale changes, and the real-time performance and stability were good.%提出一种基于均值漂移的自适应跟踪窗口算法.在初始时应用边缘加权概率密度的方法判断目标的变化,目标变大时采用形心定位和特征点仿射模型对跟踪窗口进行更新,目标减小或不变时通过Bhattacharyya系数来决定目标缩小的比例.实验表明该算法能够有效地跟踪尺度变化的目标,同时具有很好的实时性、稳定性.