基于粒子滤波的移动物体视觉跟踪方法

Visual tracking of moving object based on particle filter

基于颜色特征的粒子滤波算法已成为移动物体跟踪的热点.提出一种基于加权颜色直方图的粒子滤波跟踪算法,利用Bhattacharyya距离来描述粒子与目标区域颜色模型的相似性.实验结果表明:该方法具有较好的实时性与鲁棒性,可应用在视频监控、小车寻迹等场合.

The particle filter algorithm based on color feature has been a research focus in tracking of moving ob- ject. This paper presents a particle filter algorithm based on a weighted color histogram to track moving object, using Bhattacharyya distance to describe the color similarity between the particle and the target area, which improves the measurement matching and sample weight updating. The similarity between target template and particle area is cal- culated, which can locate the target accurately and efficiently. The implementation of this method produces robust results for different situations such as partial occlusion, rotation and shape distortion. The results show that the method is feasible and can perform well in video surveillance and vehicle tracking.