Visual tracking enhancement by trajectory simulation based on hidden semi-Markov model

Abstract

In this Letter, the authors present a novel tracking system, in which tracking accuracy can be enhanced by trajectory simulation. They generate synthetic trajectories based on observed trajectories by adopting the hidden semi-Markov model (HSMM). In the course of trajectory simulation, HSMM can encode representative states and speeds of the targets. The simulated trajectories enforce the proposed tracker to focus on the areas where targets will move at following frames. Experimental results demonstrate that it is easy to integrate the proposed trajectory simulation into traditional visual tracking methods and the trajectory simulation can considerably improve the accuracy of visual trackers.