Tracking with split and merge processes

In this paper, We propose a novel algorithm to reconstruct particle filter trackers automatically using split and merge technology. In the split process, the tracker splits itself into two or more trackers to deal with complicated and inconstant environments. In the merge process, the best one is selected from the trackers constructed in the split process, as a result the computation cost is reduced by merging useless trackers. We propose three split criteria in split process to reduce target lost probability and perform a valid split. With split and merge processes, our algorithm achieves good tracking results even using fewer particles; furthermore, as using fewer particles in our algorithm, the tracker with split and merge processes is more efficient than the standard tracker. Experiments are provided to demonstrate that the performance of the proposed algorithm outperforms that of the traditional tracker without split and merge processes.