Multi-Sensor Tracking of Move-Stop-Move Targets

This paper considers targets that exhibit move-stop-move motion patterns, for which we have kinematic-sensor detections. Our particular focus is how best to score the association of tracklets that result from a first stage of single-sensor kinematic tracking. The tracklets may originate from one or more sensors, and target-emission tracks may be available that typically exhibit low measurement rates compared to kinematic sensors. We propose a tracklet-association scoring approach that relies on hybrid-state target modeling and the maximum a posteriori target mode sequence. We describe the use of this filter as part of a graph-based tracking approach that reduces the complexity associated with the multiple-hypothesis tracking solution.