Tracking by auto-reconstructing particle filter trackers

We propose a novel algorithm named ARPF (Auto-Reconstructing Particle Filter) which reconstructs particle filter trackers automatically using split and merge technology. The algorithm deals with complicated and inconstant environments by splitting the tracker into two or more ones. In the merge process the best one is selected from the trackers constructed in the split process, and as a result the computation cost is reduced by merging useless trackers. With split and merge, the algorithm can get good tracking results even using fewer particles. The trackers constructed in the split process can track a target from different positions and directions, and therefore can reduce the probability of losing the target. Compared with other methods, the experimental results of our ARPF method show better effectiveness and efficiency.