This paper proposes a fast method for Synthetic Aperture Radar (SAR) - based target identification. First, high range resolution (HRR) profiles are extracted from the SAR data, to be used in a fast, one-dimensional (1D) HRR template-matching classifier. The role of this classifier is to prune the hypothesis space by producing a set of “best guesses.” Then, a two-dimensional (2D) SAR template classifier makes a final decision, performing a limited number of template matches based on the HRR results. The hybrid method yields results up to six times faster than a strictly 2D template classifier, with very little sacrifice in classification accuracy.
EUSAR04 - European Conference on Synthetic Aperture Radar