Tracking “bunching” multitarget correlations

In point process theory, permanental processes are used to model statistical populations whose members tend to be attracted to each other (“bunch”). This paper initiates what appears to be the first application of permanental processes to multitarget detection and tracking. Permanental processes can be used to construct bivariate-Poisson models of statistical correlations between two Poisson multitarget populations. We introduce a recursive Bayes filter for such permanentally-correlated multitarget systems. Then, by analogy with the probability hypothesis density (PHD) filter, we derive first-order approximate filter equations. This permanental-PHD filter requires the (removable) assumption that probability of detection is unity.