一种基于颜色和边缘特征的Mean Shift目标跟踪算法

针对基于颜色特征的Mean Shift算法对动态变化环境适应能力差,容易丢失跟踪目标的问题,提出了一种颜色和边缘特征相结合的Mean Shift目标跟踪算法。该算法使用核直方图概率模型描述目标,通过权值的自适应变化来评估每个特征在跟踪场景中的可靠性,并依据权值提出一种选择性模板更新方法。实验结果表明,提出的算法不仅可以克服目标在跟踪过程中发生的模板变形,而且对目标的平移旋转以及光照变化都具有较好的鲁棒性。

Mean Shift target tracking algorithm based on colour and edge features

Mean Shift algorithm based on color feature is liable to lose tracking targets,due to its poor ability to adapt to dynamic changing environment.To solve the problem,a new Mean Shift target tracking algorithm is established based on both color and edge features.The algorithm describes object features by using kernel histogram probabilistic model,evaluates the reliability of each feature in the tracking scenes through the adaptive changes of weights,and gives a selective template updating method according to the weights.Experimenttal results show the proposed algorithm can not only conquer template deformation that may occur in the tracking process,but also has better robustness towords translation,rotation and illumination changes.