An interacting multiple model-joint belief probabilistic data association approach to group tracking.

摘要

The evolution of target tracking has resulted in many algorithms suited for various multitarget tracking applications; however no algorithms provide a suitable implementation for group tracking. In advancing the tracking field, one issue that plagues the implementation of multitarget tracking applications is that presenting numerous targets on a display exceeds the human's ability to follow the tracks. To facilitate the implementation of tracking algorithms for the user, we seek to group the targets before presenting the track solution to a display to minimize clutter, enhance algorithmic efficiency, and provide a sustainable situational awareness. The GRoup IMM Tracking (GRIT) algorithm combines the Joint Belief-Probability Data Association (JBPDA) approach with an Interactive Multiple Model (IMM) estimator to track, identify, and group multiple moving targets. A Multiple Validation Gate Model (MVGM) is introduced that captures the behavior of highly maneuvering targets through move-stop-move cycles. A Targets Constrained on Road Network (TCORN) feature is also incorporated that assists with the elimination of clutter measurements that are located ‘off-road’. The GRIT algorithm additionally realizes many benefits that result from the utilization of group information feedback to accommodate many of the problems associated with multitarget tracking. This group information feedback is shown to enhance the tracking of merging and splitting tracks within groups and targets engaging in move-stop-move maneuvers. The GRIT algorithm also facilitates the tracking of occluded targets and provides for the capability to autonomously initiate new groups and terminate the track of multiple groups.*; *This dissertation includes a CD that is compound (contains both a paper copy and a CD as part of the dissertation). The CD requires the following application: Microsoft Office.