Probabilistic multi-hypothesis tracker: addressing some basic issues

Abstract

For some years, the authors and other researchers have sought to improve the PMHT, to increase its elegance or to show its superiority in terms of lost tracks. While all of these are important, it must be admitted that the PMHT is unlikely to gain much practical acceptance until a number of practical ‘pedestrian’ issues are addressed. There was considerable progress by Luginbuhl et al. (2001), in which a track-management system (track-initiation, -termination and -merging) was proposed. The authors continue in that vein: how does the PMHT deal with nonlinear models, with consistency checks and with out-of-sequence measurements?