A comparison of track initiation with PDAF and PMHT

A problem with many algorithms for target tracking is that they are designed to update tracks under the assumption that the tracks correspond to real targets, and that the initial conditions of the targets are known. In practice, this is not the case, and practical algorithms must be capable of track initiation and termination. One method of solving these problems is to introduce a Markov model to estimate the validity of each track. This approach has been referred to as target visibility. This paper compares the single target track initiation performance of the probabilistic data association filter (PDAF) with the probabilistic multi-hypothesis tracker (PMHT) when each uses the visibility target model.