High-quality interferometric inverse synthetic aperture radar imaging using deep convolutional networks

摘要

In this article, a modified complex-valued convolutional neural network (MCV-CNN) specifically for interferometric inverse synthetic aperture radar (InISAR) imaging is proposed. Comparing with the fast Fourier transformation-based and sparsity-driven imaging algorithms, the MCV-CNN can achieve super-resolution and side-lobe suppression on the imaging results simultaneously within a short time. The inputs of the MCV-CNN are complex-valued radar echo data, and the outputs are complex-valued ISAR images which contain both the amplitude and phase information. Then the phase information is adopted to perform an interferometric operation, and the high-quality three-dimensional InISAR imaging results can be achieved. A 0.22 THz InISAR imaging experiment has been carried out to show the superiority of the proposed method on imaging quality and computational efficiency.