On track-management within the PMHT framework

Tracking multiple targets in a cluttered environment is a challenging task. Probabilistic multiple hypothesis tracking (PMHT) is an efficient approach for dealing with it. Essentially, PMHT is based on expectation-maximization for handling association conflicts. Linearity in the number of targets and measurements is the main motivation for a further development and extension of this methodology. However, as PMHT is a target-oriented approach, it requires the number of targets in the surveillance area to be a priori known. Furthermore it assumes that all tracks are true tracks. Of course, this demand cannot be met in a realistic scenario and drastically limits the usage of PMHT in practical applications. To achieve a complete solution for automatic multiple target tracking in clutter, the PMHT framework has to be supplemented by a track management system (TMS). In this paper, we address the most important issues of a TMS. Thereby we focus in particular on the problem of track deletion. For this purpose, we consider two likelihoodratio (LR) approximations that can be calculated as a by-product of the PMHT iteration process. The first one is a Generalized LR approach and the second one is based on synthetic measurements. Both approaches are evaluated in combination with the wellknown Page test and the Shiryaev test, and are compared with a particle filter approximation.