Three-dimensional SAR target reconstruction based on pix2pix and pixel2mesh

Tomographic synthetic aperture radar (SAR) is one type of technology that can achieve three-dimensional (3D) high-resolution imaging of targets. However, it requires collecting and processing a large amount of data each time, which is time-consuming and costly. The 3D reconstruction technology based on deep learning can directly reconstruct targets from two-dimensional (2D) SAR images, requiring only sufficient data to train the network. In this paper, a 3D reconstruction network which combines pixel-to-pixel (pix2pix) with pixel-to-mesh (pixel2mesh) is proposed to reconstruct 3D SAR targets from SAR images. The pix2pix is used to generate an optical image corresponding to the input 2D SAR image, and then the pixel2mesh is used to complete the 3D reconstruction from the generated 2D optical image. Experimental results on the simulated and real SAR datasets verify that the combined network achieves good 3D reconstruction results.