基于区域协方差矩阵的模板更新方法分析与比较

Analysis and Comparison of Template Update Method Based on Region Covariance Matrix

在目标跟踪算法中,目标特征描述子和模板更新策略都是跟踪成败的关键。本文在选用区域协方差矩阵描述子的基础上,结合粒子滤波器算法,并且采用两种常用的模板更新方法,分别对商场通道监控视频序列中的行人进行了跟踪仿真实验。通过仿真分析并比较不同模板更新方法条件下的跟踪性能,从而在不同情况下选择合适的模板更新方法,以得到更精确的跟踪结果。

The target feature descriptor and template update strategy are the keys to success in target tracking algorithm. In the paper, we used two common template updating methods base on region covariance matrix descriptor to track the people in shopping malls channel surveiliance video sequences by using particle filter algorithm respectively. By means of simulation, we analyzed and compared the tracking performances by using different template updating methods, and then selected the appropriate template updating method to acquire the more accurate tracking results under different situations.