An extended unifying approach to multi-target tracking

Abstract

By means of global modeling of multitarget tracking (MTT), data-association and maneuver-estimation problems can be simultaneously solved using system identification techniques. With this approach, previously developed single-target tracking-acceleration estimation techniques can also be directly used for the MTT problem. In particular multiple-model (adaptive) Kalman filtering (MMKF) can be used to obtain the optimal solution. For computational considerations, one can apply some suboptimal solutions of MMKF such as one-step conditional maximum likelihood or maximum posteriori estimation.<>