New multi-step sampling with adaptive sampling patterns in particle filtering for tracking in surveillance systems

Particle filtering is one of the most efficient approaches for object tracking in video application systems. In this paper, we propose a new multi-step recursive sampling method to replace the conventional direct importance sampling. An online-adaptive sampling pattern for proposal distributions is established. New particles are then sampled recursively from the existing particles with high weights. A 2D predictive transition vector is used to update the pattern of the multivariate Gaussian sampling. Experimental results illustrate that the proposed method reduces computation substantially and it also preserves good tracking results comparable to other algorithms in the literature.