Track-Oriented MHT with Unresolved Measurements

This paper first validates that the track-oriented multiple-hypothesis tracking recursion holds in the case of state-dependent detection probabilities, as is generally assumed. Next, we seek to extend the track-oriented multiple-hypothesis tracking recursion to allow for unresolved measurements. The formulation requires some simplifying assumptions, including an assumption that targets be resolved at birth and a restriction on the size of unresolved target clusters. The tracking recursion requires some approximation to admit track-oriented (factored) form, and leads to a nonlinear optimization problem. We discuss a multi-stage architecture that provides a simpler and more robust processing approach for practical settings.