Clustering and Association for Active Sonar Tracking and Classification -- 19-014

We propose to leverage and extend a new and innovation approach to automate contact clustering/association and tracking of extended target highlights based on the JPDA Intensity Filter (JiFi) [Str10]. JiFi is a hybrid tracking system that combines aspects of the Joint Probabilistic Data Association (JPDA) filter and the multitarget intensity filter (iFilter). Both are Bayesian filters and both produce estimates of a posterior distribution. They estimate different functions that are complementary. The iFilter estimates a distribution on the number and locations of targets, and the JPDA filter estimates a distribution on target locations, conditional on the number of targets. More specifically, the iFilter estimates the multitarget intensity function, which is the expected number of targets per unit state space, and the JPDA conditions on the number of targets and estimates a probability density function for each target. The JiFi filter uses the iFilter as a surrogate for the prior distribution for target highlights, not the actual targets themselves.