A new real-time robust object tracking method

Abstract

In this paper, we propose a real-time robust object tracking method that is based on a generative visual appearance model but with some level of background awareness. We introduce several strategies to ensure a stable visual appearance model for the target as tracking progresses. The strategies include modeling of a foreground color distribution, maintaining a set of foreground templates with the target region emphasized, and incorporating background templates into the dictionary for sparse representation of the target appearance. Experimental results demonstrate that our method outperforms several latest state-of-the-art tracking methods in terms of tracking performance.