False alarm reduction through road area delimitation

Abstract

Proposes the concept of road area delimitation to improve the performance of a system that detects ground-based military vehicles in synthetic aperture radar (SAR) imagery. Road area delimitation is a data-reduction procedure that consists of automatically finding roads in the imagery, and subsequently discarding the imagery that is not within some prespecified distance of a road. This initial data-processing step eliminates all false alarms in the discarded region, thereby yielding improved detection performance for cases in which vehicle deployment and movement are restricted to the area near roads. The authors develop an algorithm whose function is to estimate road locations automatically, and show the results of applying this algorithm to actual SAR imagery.