Super-Resolution for MIMO Array SAR 3-D Imaging Based on Compressive Sensing and Deep Neural Network

Abstract

Multiple-input multiple-output (MIMO) array synthetic aperture radar (SAR) can straightly obtain the 3-D imagery of the illuminated scene with the single-pass flight. Generally, the Rayleigh resolution of the elevation direction is unacceptable due to the length limitation of linear array. The super-resolution imaging algorithms within the compressive sensing (CS) framework have been extensively studied because of the essential spatial sparsity in the elevation direction. However, the super-resolution performance of the existing sparse reconstruction algorithms will deteriorate dramatically in the case of lower signal-to-noise ratio (SNR) level or a few antenna elements. To overcome this problem, a new super-resolution imaging structure based on CS and deep neural network (DNN) for MIMO array SAR is proposed in this article. In this new algorithm, the spatial filtering based on CS is first proposed to reserve the signals only impinging from the prespecified space subregions. Thereafter, a group of parallel end-to-end DNN regression models are designed for mapping the potential sparse recovery mathematical model and further locating the true scatterers in the elevation direction. Finally, extensive simulations and airborne MIMO array SAR experiments are investigated to validate that the proposed method can realize the state-of-the-art super-resolution imaging against other existing related methods.