A Compressive Sensing Approach to Multistatic Radar Change Imaging

This paper describes a new approach for forming change images from multistatic radar data based on compressive sensing (CS). Broadly speaking, change images are naturally sparse in the raw image domain, which suggests a CS reconstruction method. Recent results show that the sparsity of the estimand dictates the number of samples required for faithful reconstruction, meaning a change image can be formed with far fewer measurements than used for conventional radar imaging. Our application has a small number of antennas arranged around the perimeter of a surveillance region, which provide large angular diversity but very poor angular sampling. Furthermore, due to application constraints, the scene is interrogated with limited frequency diversity. We aim to construct a high-resolution change image from the measurements, which are sub-Nyquist both spatially and in frequency. This paper first develops a radar imaging model in the context of CS, and then shows with collected data that a sparseness model improves image utility over conventional methods in our setting.