New pruning/merging algorithm for MHT multitarget tracking

In this paper we develop a new general Viterbi algorithm for multi-target tracking. A standard Multiple Hypothesis Tracking (MHT) formulation, based on Maximum A Posterior (MAP) estimation, is considered for optimally associating measurement data over time to form estimates of the multiple tracks. For estimating K tracks, a trellis diagram of the measurements is used to depict all track set hypotheses as trellis paths. MAP path costs are computed using Kalman filters and a priori track set probabilities. A new K-track, List-Viterbi algorithm is then employed to effectively prune the number of evaluated hypotheses to a manageable level. Further hypothesis reduction is implemented through path merging and trellis truncation. The resulting Viterbi MHT algorithm is sequential. It is very flexible in that it can handle missed detections, false alarms and number-of-track estimation. It provides an ordered list of best track sets, which can be useful in subsequent data fusion. Compared to MHT algorithms which prune candidate track sets using `gating volumes', the new Viterbi MHT is less prone to loosing tracks.