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 paper proposes a deep unrolling network named NSR-NET, which is based on SAR image representation learning and is applicable for undersampled imaging in non-sparse 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 2D 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 non-sparse 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.

Various imaging methods based on compressed sensing (CS) of synthetic aperture radar (SAR) have been proposed to reduce the sample size of echoes required for the imaging process. The unrolling technique further solves the inefficiency of conventional CS-based methods by mapping them into deep neural networks. However, most of these methods are based on sparsity prior of the scene or its transformation domain, which could be invalid for non-sparse scenes. To address this, we proposed a network utilizing the feature priors of the images instead of sparsity for non-sparse scene reconstruction of SAR, namely NSR-Net. We adopt learnable regularization terms in the CS model. Then the iterative solving process of the model is derived and unrolled into the proposed deep neural network to learn the best regularization terms from data. Simulation experiments verified the effectiveness of NSR-Net in the reconstruction of non-sparse scenes with down-sampled SAR echoes.

Synthetic aperture radar (SAR) can obtain two-dimensional images of the illuminated area, which is an important means for earth remote sensing and monitoring. However, due to the loss of azimuth data and system errors during the processing of data sampling, it is necessary to study the method for high-quality SAR image reconstruction from down-sampled data in the condition of measurement inaccuracy. Considering these factors, this paper proposes a sparsity-driven SAR imaging method based on general regularization and the sparse total least-squares (S-TLS) model and implements the method by an unfolded deep network. In the proposed method, general regularization can solve the problem of sparse sampling, and the S-TLS model is adopted to deal with measurement inaccuracy. Moreover, through the deep network implementation, the proposed is more time-efficient and can exploit more effective scene prior knowledge, making the proposed method suitable in practical applications. Experiments verify the effectiveness of the proposed method.

Synthetic aperture radar (SAR) imaging with sub-Nyquist sampled echo is a challenging task. Compressed sensing (CS) has been widely applied in this case to reconstruct the unambiguous image. The CS-based methods need to set the iterative parameters manually, but the appropriate parameters are usually difficult to obtain. Besides, such methods require a large number of iterations to obtain satisfactory results, which seriously restricts their practical applications. Moreover, the observation scene of SAR is not sparse in some cases. In this paper, we aim at proposing an efficient and effective imaging method for non-sparse observation scenes with reduced data. Firstly, considering the characteristics of non-sparse observation scenes in SAR imaging, we model the SAR imaging problem as a joint low-rank and sparse matrices recovery problem. After that, the iterative alternating direction method of multipliers (ADMM) to solve the above problem is unrolled into a layer-fixed deep neural network with trainable parameters, in which the learnable parameters are layer-varied. The threshold parameters, as well as the weight parameter between the sparse part and low-rank part of each layer, are learned adaptively instead of manually tuned. Experiments prove that the proposed LRSR-ADMM-Net is capable of reconstructing the non-sparse observed scene with high efficiency and precision. Particularly, the proposed LRSR-ADMM-Net yields better reconstruction performance while maintaining high computational efficiency compared with the state-of-the-art iterative recovery methods and the trainable sparse-based network methods.

Synthetic aperture radar (SAR) is an important means for target surveillance through reconstructing the microwave image of the observation area. However, under the condition of low signal-to-clutter ratio (SCR), such as a strong sea clutter situation, it is difficult to surveil targets from SAR images acquired by the traditional matched filter-based imaging methods. To improve the target surveillance performance of SAR, this article proposes a target-oriented SAR imaging method, which can enhance the desired target and improve the SCR in the reconstructed SAR images. By separating the target area from the clutter area, we first establish a target-oriented SAR imaging model, where the generalized regularization is used to characterize the features of the target, contributing to the improvement of SCR in the reconstructed image. Then, the imaging model is solved through a deep network, MF-ADMM-Net, which is obtained by unfolding an alternating direction method of multipliers (ADMM)-based iterative solution. In addition, the training strategy is formulated with the consideration of complex values. Experiments are conducted to verify the performance of image reconstruction and SCR improvement of the proposed method, and comparisons show the superiority of MF-ADMM-Net in effect and efficiency.