Appearance Model-Based Particle Filter for Robust Tracking

Peng Tian,Zhou Yue

DOI: https://doi.org/10.3969/j.issn.1007-757X.2010.11.014

2010-01-01

Abstract:In this paper,an appearance model-based particle filter was proposed for targets tracking in visual surveillance,which means the particle filter use the proposed appearance model,as the description measure of the target's visual feature,to built the target template.Being different from the color histogram,the proposed appearance model not only has the target's color information,but also retains its pixels' spatiality information.Even if going through a long process,the appearance model,owing to its continuous update,is still effective to be the reference template for the candidate target of the current frame.Based on the above advantages,the proposed tracking algorithm can robustly track the target in the scenes,where the target has changed a lot in its size,position and rotation,or where the target is occluded by the background.In the multi-targets tracking applications,utilizing the depth technique,the proposed tracking algorithm can effectively handle the intractable problem-occlusion arises between targets.Comparing with the color-based particle filter for targets tracking,the experimental results fully demonstrate the robust performance of our method.