Mask particle filter for similar objects tracking

Abstract

Tracking appearance similar objects is very challenging. Conventional approaches often encounter "hijack" problem. That is to say, the tracking results for the smaller objects will be attracted to the larger one in the close vicinity. In this paper, we propose a decentralized particle filter approach for similar objects tracking. When the objects are close, the tracking results for the larger one will be masked and its influence will be eliminated. In principle, the tracker for the smaller object needs to be run two times, which increase the time costs. To tackle this, we construct the integral image for the mask region and dramatically decrease the calculation time of the evaluation of likelihood functions in the masked image. Experimental results show that the proposed approach effectively avoids "hijack" problems. © 2009 Springer Berlin Heidelberg.