Track before detect DOA tracking of extended targets with marked poisson point processes

In this paper we propose a novel Track Before Detect (TBD) filter aimed at tracking multiple extended targets from phased-array observations. For extended targets, the source signal is angularly distributed, and hence we track the centroid Direction Of Arrival (DOA) of the target generated signal - called target signal. In this work we suppose known the shape and extent of the target-signal angular spread. Solutions based on extending the system state, to include the target signal, lead to higher-dimensional posteriors. We avoid an extended state by using a novel Marked Poisson Point Process (MPPP) model for the system, and accordingly, we derive the intensity/PHD filter that adaptively estimates target number and corresponding centroid DOAs. The source signals are interpreted as the mark of a target, and they are analytically integrated in the update formula of the filter. Therefore, an efficient particle filter implementation is possible. Results on simulated data showcase the improved results of the proposed filter over state-of-the-art methods.