Detection of broadside targets during image formation using a quadtree approach

Abstract

The military is interested in using ultra-wideband (UWB) synthetic aperture radar (SAR) systems to detect ground targets. Standard automatic target detection methods search the entire scene for regions of interest (ROI) after image formation. In order to save computations, we introduce a multiscale detection algorithm that uses partially processed radar data during the intermediate stages of a quadtree-based backprojection image formation algorithm. When the detector accrues enough information to determine that a patch of ground is free of potential targets, it then cues the image former to terminate the processing that would further resolve that patch. The detector combines a feature that estimates the coherent signal to noise ratio with another feature that exploits the "broadside flash" scattering phenomenon. The new approach is evaluated over a measured database generated by the ARL BoomSAR radar.