Tracking Dim Target in Infrared Imagery Using the Trust Region Embedded Particle Filter
Zhenyu Wang, Kedai Zhang, Yi Wu, Hanqing Lu · 2006
This paper propose a novel algorithm, the trust region embedded particle filter (TREPF), for target tracking in infrared imagery. Trust regions and particle filters are two successful methods for object tracking. The presented TREPF algorithm integrates the advantages of the two approaches. Contrasting the original particle filters and trust regions, the new algorithm can maintain multiple hypotheses with fewer particles by encouraging the particles to be in the right part of the state space. Because of these properties, the sample degeneracy problem and the sample impoverishment problem are overcome while the computational cost is reduced. Promising experimental results on several infrared sequences demonstrate the robustness and effectiveness of TREPF