XMAP: Track-to-track association with metric, feature, and target-type data

The Extended Maximum A Posteriori Probability (XMAP) method for track-to-track association is based on a formal, Bayesian methodology for incorporating metric, feature, and target-type data. The metric component improves upon the classical derivation of the adaptive threshold to produce a more robust alternative, which can handle clusters with very few tracks and tracks with large covariances. The feature and target-type components are treated jointly, allowing for the possibility that the performance of the feature extractor depends on target type. This coupling allows feature information to be interpreted differently depending on the results of a target classifier-from a feature measurement being deemed accurate within within a small tolerance, to the measurement being thrown out altogether. A key innovation in the derivation is the non-informative noise assumption used in the feature measurement model, which gives a simple, robust form to the results.