Abstract: The traditional particle filter tracking algorithms is only based on color feature. And under complex background or shelter with color closing to targets’, particle filter tracking algorithm can almost fail. Particle filter based on multi-feature fusion fuses targets’ color characteristics and edge feature according to certain weighting coefficient to build a likelihood function as a criterion for tracking targets. It overcomes the shortage of tracking algorithm based on color feature only. Experimental results show that multi-feature particle filter tracking algorithm has good performance of tracking.

Key words: particle filter, multi-feature fusion, edge feature, target tracking

摘要： 传统的粒子滤波跟踪算法只依靠单一的颜色特征作为跟踪依据。在复杂背景或者遮挡物颜色与跟踪目标接近时传统的粒子滤波算法很容易造成跟踪目标丢失。针对该问题，提出一种基于多特征融合的粒子滤波算法，该算法按一定的权值系数利用目标的颜色特征和边缘特征来构建似然函数作为跟踪目标的跟踪依据，克服了依靠单一颜色特征跟踪目标的跟踪算法的不足。实验结果表明，多特征融合后的跟踪算法有较好的跟踪性能。

关键词: 粒子滤波, 多特征融合, 边缘特征, 目标跟踪

CLC Number: 

• TP391.41