Tracking Targets with Specified Spectra Using the H-PMHT Algorithm

The H-PMHT algorithm is applied to sensor-level, multi-target tracking problems in which target and noise power spectra are assumed to be specified a priori. The resulting algorithm, based on the expectation-maximization method, is equivalent to a bank of iteratively reweighted smoothing filters. These results constitute a potentially important algorithm for tracking multiple targets in hyperspectral Image sequences.