Target indexing in synthetic aperture radar imagery using topographic features

Automatic target indexing using synthetic aperture radar (SAR) imagery is investigated. Target areas are located using a constant false alarm rate (CFAR) detector. Classification of targets is performed using the topographical primal sketch, which assigns each pixel a label that is invariant under monotonic gray tone transformations. The use of a topographic scheme based on the curvature of the underlying image intensity allows for the classification of targets in high resolution imagery. Target matching is performed using correlation and graph matching schemes that incorporate quantitative values associated with each feature as well as qualitative feature labels. Experimental results for indexing of targets performed at 1 foot and 1 meter resolutions are presented