Distributed MHT and Extensions

Despite the proven success of Multiple Hypothesis Tracking (MHT) as a methodology for Multi-Target Tracking (MTT), computational constraints and other fundamental performance limitations may lead to unacceptable performance in some settings. This lecture focuses on the benefits that can be achieved with multi-stage MHT processing. In many settings, judicious distributed MHT processing enables improved performance over (necessarily suboptimal) centralized MHT. We provide illustrative examples from several domains. Additionally, we describe recent advances in graph-based tracking, a fast (approximate) approach to MHT that provides improved results in certain applications.