A generalised labelled multi-Bernoulli filter for extended multi-target tracking

Abstract

This paper addresses extended multi-target tracking in clutter, i.e. tracking targets that may produce more than one measurement on each scan. We propose a new algorithm for solving this problem, that is capable of initiating and maintaining labelled estimates of the target kinematics, measurement rates and extents. Our proposed technique is based on modelling the multi-target state as a generalised labelled multi-Bernoulli (GLMB), combined with the gamma Gaussian inverse Wishart (GGIW) distribution for a single extended target. Previously, probability hypothesis density (PHD) and cardinalised PHD (CPHD) filters based on GGIW mixtures have been proposed to solve the extended target tracking problem. Although these are computationally cheaper, they involve significant approximations, as well as lacking the ability to maintain target tracks over time. Here, we compare our proposed GLMB-based approach to the extended target PHD/CPHD filters, and show that the GLMB has improved performance.