Abstract:Synthetic aperture radar (SAR) is widely used in various fields due to its all-weather and all-day working characteristics. With the increasing use of SAR on small platforms, SAR is facing a series of problems due to the large volume of echo data. Imaging methods based on compressed sensing (CS) use the sparsity prior of the scene to reconstruct images from undersampled echoes. However, the CS-based method requires the imaging scene or its transformation domain to be sparse, which is not the case for most practical applications. This article proposes a deep unrolling network named NSR-NET, which is based on SAR image representation learning and is applicable for undersampled imaging in nonsparse scenes. In modeling, the learned image representation is adopted as the regularization term. Then, the proximal gradient descent (PGD) algorithm was used to derive the iterative solution of the model. In network design, the iterative process is unrolled into a deep neural network with learnable parameters. Specifically, image representation is obtained through 2-D convolutional layers in the network, and a learnable piecewise linear layer is used to fit the regularization function, which ultimately achieves the mapping from undersampled echoes to SAR images. Comparative experiment using different imaging methods shows that the imaging performance of the proposed network exceeds that of the state-of-the-art methods in nonsparse scenes. Moreover, we also designed transferability validation experiments with different radar parameters and imaging scenes, whose experimental results suggest that the proposed network has good generalization ability.

A Deep Learning Method for SAR Imaging in Sparse Field

SAR Fast Target Imaging in Sparse Field Based on AlexNet

A New SAR Imaging Method for Sparse Field Traget

DeepRED Based Sparse SAR Imaging

End-to-End SAR Deep Learning Imaging Method Based on Sparse Optimization

Learning-Based Sparse Recovery Algorithm for 3D SAR Imaging

Deep Learning Meets SAR: Concepts, models, pitfalls, and perspectives

SAR Nonsparse Scene Reconstruction Network via Image Feature Representation Learning

SAR-NeRF: Neural Radiance Fields for Synthetic Aperture Radar Multi-View Representation

A Deep Learning Approach for SAR Tomographic Imaging of Forested Areas

Deep SAR Tomography: A Model-Inspired Approach With Learned Sparse Regularizer

RDAnet: A Deep Learning Based Approach for Synthetic Aperture Radar Image Formation

Fast Super-resolution 3D SAR Imaging Using an Unfolded Deep Network

An Advanced SAR Image Despeckling Method by Bernoulli-Sampling-Based Self-Supervised Deep Learning

Near-filed SAR Image Restoration with Deep Learning Inverse Technique: A Preliminary Study

SAR Neural Radiance Fields with Multi-Resolution Hash Encoding

From Theory to Application: Real-Time Sparse SAR Imaging

Nonsparse SAR Scene Imaging Network Based on Sparse Representation and Approximate Observations

Detection of Oil Spill Using SAR Imagery Based on AlexNet Model

Deep Learning as Applied in SAR Target Recognition and Terrain Classification

Self-supervised Enhanced Radar Imaging Based on Deep-Learning-Assisted Compressed Sensing