Tracking multiple extended objects - A Markov chain Monte Carlo approach

This paper proposes a multiple object tracking system for spatially extended objects, whose number is a priori not known and dynamically changing over time. Compared to the expected size of the objects, a high resolution range measuring sensor is used within an implementation of the proposed system. For that, the Bayesian framework is rigorously utilized and implemented using a reversible jump Markov chain Monte Carlo sampling approach. A priori knowledge like object dynamics is statistically expressed and integrated into one Bayes filter. This includes how objects lookalike and move, where they are expected to appear & disappear, and how they do interact with each other. The functionality of the system is shown in simulative results.