Abstract:As high-frame-rate synthetic aperture radar (SAR) has the ability to form continuous SAR images and dynamically monitor the ground areas of interest, it has attracted more and more attention nowadays. In practical applications, the enormous data in high-frame-rate SAR system to obtain the multiframe images brings big challenges to its transmission, storage, and processing. In order to solve this problem, there are many recent papers on formulating the high-frame-rate SAR imaging problem into a low-rank tensor recovery problem, and correspondingly, the sampling amount of the high-frame-rate SAR data can be largely reduced. However, existing algorithms to solve the low-rank tensor recovery problem in high-frame-rate SAR application still suffer from large computational cost. Under the above inspiration, this article proposes a deep neural network architecture for high-frame-rate SAR imaging, i.e., tensor alternating direction method of multiplier network (TADMM-Net), which is more computationally efficient in the imaging procedure. Specifically, we formulate the high-frame-rate SAR imaging processing into a low-tubal-rank tensor recovery problem. We solve the low-tubal-rank tensor recovery problem using a tensor alternating direction method of multiplier (ADMM) algorithm and then design a new deep neural network architecture by applying algorithm unfolding techniques to the underlying low-rank tensor recovery problem. The proposed TADMM-Net approach shifts the computational burden from the testing phase to the training phase, and the practical processing time can be extremely decreased compared with the existing algorithms for the low-rank tensor recovery problem in high-frame-rate SAR imaging applications. It also offers various advantages over the existing high-frame-rate SAR imaging algorithms, including higher performance in the case with low sampling amount, lower storage complexity, and no requirement for handcraft hyperparameters adjustment. The methodology was tested on high-frame-rate SAR data. These tests show that the proposed architecture outperforms other state-of-the-art methods in high-frame-rate SAR imaging applications.

ORTP: A Video SAR Imaging Algorithm Based on Low-Tubal-Rank Tensor Recovery

Video SAR Imaging Based on Low-Rank Tensor Recovery

Joint Low-Rank and Sparse Tensors Recovery for Video Synthetic Aperture Radar Imaging

A TLRTV Dual-band SAR Image Denoise-Fusion Strategy and Its Preliminary Experiment Analysis in Multi-band Airborne Radar System

Video Formation Method for UAV SAR Utilizing Tensor Recovery Algorithm.

Video SAR Reconstruction Based on Low-Rank Representation

SAR Image Reconstruction and Target Extraction with Under-Sampled Data Via Low-Rank and Sparsity Matrix Decomposition.

OSRanP: A Novel Way for Radar Imaging Utilizing Joint Sparsity and Low-Rankness

NPTT: Nonlinear Spatial-Temporal Transform for Low-Rank Tensor Recovery in Infrared Small Target Detection.

Online Robust Low-Rank Tensor Modeling for Streaming Data Analysis

Learning-Based High-Frame-Rate SAR Imaging

Robust low tubal rank tensor recovery via L 2 E criterion

Low rank plus sparse decomposition of synthetic aperture radar data for target imaging and tracking

Tensor Recovery via Nonconvex Low-Rank Approximation

Fast Dynamic ISAR Imaging Method Based on Low-rank Tensor Decomposition with Alternating Minimization.

Tensor Recovery with Weighted Tensor Average Rank.

Dynamic ISAR imaging of maneuvering targets based on sparse matrix recovery

Infrared Small Target Detection via Nonconvex Tensor Fibered Rank Approximation

Simultaneous Rectification and Alignment via Robust Recovery of Low-rank Tensors.

Infrared Cirrus Detection Using Non-Convex Rank Surrogates for Spatial-Temporal Tensor

A Joint Moving Target Detection Method in Video SAR Via Low-Rank Sparse Decomposition and Transformer