Polarimetric classification of scattering centers using M-ary Bayesian decision rules

A Bayes-optimal decision rule is presented for detection and classification of scattering centers in clutter. Scattering centers are modeled as one of M canonical reflectors with unknown amplitude, phase and orientation angle; clutter is modeled as a spherically invariant random vector. A choice of costs in the Bayes risk is shown to yield a two-stage classification rule. The first stage is a Neyman-Pearson detector which rejects clutter, whereas the second stage classifies the detection in one of the M target classes. The resulting decision rule yields computationally simple implementation, intuitive geometric interpretation, and posterior estimation of decision uncertainty. Performance of the proposed classifier is illustrated on imagery from an airborne UHF-hand radar.