Three-D Imaging of Urban Layover Scatterers Based on Learned LISTA Algorithm

This paper explores a Tomographic Synthetic Aperture Radar (TomoSAR) learning imaging model based on the Learned ISTA (LISTA) algorithm. It addresses the limitations of traditional sparse regularization methods encountered in TomoSAR, particularly their sensitivity to noise and high computational complexity. This study involves unfolding the Iterative Shrinkage Thresholding Algorithm (ISTA) into a neural network structure, allowing for the transformation of preset hyperparameters into learnable parameters, thereby enhancing model adaptability. This paper includes constructing a diverse training set and testing the model against various conditions, including different Signal-to-Noise Ratios (SNRs) and scatterer spacings. The results demonstrate that the LISTA-based TomoSAR imaging algorithm achieves super-resolution performance and reduces computational complexity over traditional methods.