Compressive sensing techniques applied to multi-look ISAR images

Abstract

Classical ISAR imaging usually is based on the polar re-formatting algorithm making use of the fast Fourier transform. If the observed aspect angle change is large enough, several partial segments can be processed separately, the resulting partial images can be summed incoherently to a multi-look image with reduced noise and speckle level and exhibiting more details, since some parts of the objects can be seen only from certain aspect angles. In this paper we regard multi-look ISAR imaging based on compressive sensing techniques. We use data obtained with a turn table, which is the standard method to determine two dimensional scattering centre distributions of a target under controlled conditions. A special implementation will be discussed using compressive sensing in the range domain, in which the object reflectivity will be naturally sparse. The reconstructed reflectivity in range allows to extend virtually the bandwidth leading to an improved resolution. Polar re-formatting is applied in this algorithm to the extended bandwidth-data. Further, the migration of specular points over the viewing angle is analysed and serves as a model for block-sparse recovery. The processing is demonstrated at real turntable data obtained with the FMCW millimeter wave COBRA radar from Fraunhofer FHR.