MRF-Particle filters for Tracking Multiple Object

This paper proposes a multi-target tracker with improved MRF-Particle filters used to solve hijacking problem; independent particle filter for tracking each ob-ject can be kidnapped by a neighboring target which has higher likelihood than real target. The motion model built by Markov random field (MRF) has been use-fully applied for avoiding being hijacked by lowering the weight of particles which are very close to any neighboring target. But it only uses pairwise term for penalty function for lowering particle’s weight. In this paper we improve the MRF-based particle filters by adapting both unary and pairwise term for observa-tion model. C. Oh, C. Lee, MRF-Particle filters for Tracking Multiple Object, 2012 International Exposition Yeosu Korea, International Conference on Information Technology (YSEC 2012), Apr. 2012