A target discrimination algorithm for high resolution SAR imagery

Target discrimination is an important stage of automatic target recognition (ATR). A region of interest (ROI) segmentation algorithm based on CFAR technique is described first. Then, a novel discriminator, the extended fractal (EF) discriminator, is proposed. At last, a discrimination algorithm composed of EF feature, target region area, and peak power ratio is designed to discriminate target from clutter ROIs. Experimental results using the MSTAR dataset show that this discrimination algorithm can eliminate most of the false alarms while maintain a high detection rate.