PHD filter with diffuse spatial prior on the birth process with applications to GM-PHD filter

This paper presents a simple and efficient way to set the birth process of the Probability Hypothesis Density filter that enhances the performance of this approach when tracking multiple targets in clutter with no a priori spatial information on where targets can appear. The novelty introduced concerns the intensity of the birth Random Finite Set that models new appearing targets. In many papers, this intensity is modelled by a Gaussian mixture whose components are “deterministically distributed” across the surveillance region where targets are more likely to appear than elsewhere. Though this assumption is valid for some specific applications, it can be too restrictive in a more general case. The simple idea underlying and motivating our approach is that targets are more likely to appear around measurements and this amounts to take a single diffuse hypothesis for the birth process.