A Multitarget Tracking Algorithm Based on Random Sets

A multitarget tracking algorithm, based on representing the set of target states and the set of measurements as realisations of finite random sets, is proposed. A measurement model allowing for missed detections and false returns and a motion model allowing for target death are defined. Target births are incorporated by combining the existing track set at the prior scan with a set of proposed birth tracks. Finite-set statistics are used to derive the global posterior density. The realisation of a practical algorithm requires that this global posterior density is approximated in some way. We propose an approximation in which a set of tentative tracks is maintained with each tentative track represented by an existence probability and a density characterising its state.