Visual object tracking is an active research field in the area of computer vision. The tracking process usually includes the construction of an object appearance model and the object localization. This paper investigates the use of Particle Swarm Optimization (PSO) as the object localization method based on the Bayesian tracking framework. The widely adopted particle filter tracking technique, however, suffers from high computational cost due to the approximation requirement of the distribution of particles. Thus, PSO is applied since it can adaptively adjust the computational expenditure according to each frame in the video. Furthermore, a new appearance model based on Earth mover's distance is proposed. The experimental results show that the proposed approach enhances the accuracy of the tracking algorithm significantly compared to the basic particle filter tracking method. Furthermore, the proposed appearance model is more robust than other Earth mover's distance based tracking algorithms.
