Particle Filter Object Tracking Based on Multiple Cues Fusion
Zhiqiang Wen, Zhaoyi Peng, Deng Xiao-jun, Shifeng Li · Procedia Engineering · 2011
In this paper, a measurement model of particle filter based on multiple cues fusion is presented. Firstly, color histogram and motion information histogram are built respectively via extracting color information and movement information, and then a weighted linear model is built as the measurement model. In addition, the center-around method is used to decide the object outline definition which is adopted to compute the weight in the weighted linear model. The parameter sigma in Gaussian function is acquired by using adaptive inverse estimation. Lastly, a mass of experiments show the presented method is effective.