A novel auxiliary particle PHD filter

We propose a novel auxiliary particle probability hypothesis density (AP-PHD) filter that elegantly combines the standard AP-filter with the particle PHD filter. The selection of particles in the proposed AP-PHD filter is based on maximizing the accuracy of the cardinality estimate. Moreover, the resampling is done on each auxiliary variable cluster separately instead of resampling particles all together without considering their different natures. Thus, from these clusters different particle sets are formed to account for detected and surviving targets, undetected but surviving targets, targets occluded and lost, newborn targets and targets reborn. Simulation results indicate that the novel AP-PHD filter improves the accuracy of both cardinality and position estimates when compared to the particle PHD filter.