Target recognition in synthetic aperture radar (SAR) imagery suffers from speckle noise and geometric distortion brought by the range-based coherent imaging mechanism. A new SAR target recognition system is proposed, using a SAR-to-optical translation network as pre-processing to enhance both automatic and manual target recognition. In the system, SAR images of targets are translated into optical by a modified conditional generative adversarial network (cGAN) whose generator with a symmetric architecture and inhomogeneous convolution kernels is designed to reduce the background clutter and edge blur of the output. After the translation, a typical convolutional neural network (CNN) classifier is exploited to recognize the target types in translated optical images automatically. For training and testing the system, a new multi-view SAR-optical dataset of aircraft targets is created. Evaluations of the translation results based on human vision and image quality assessment (IQA) methods verify the improvement of image interpretability and quality, and translated images obtain higher average accuracy than original SAR data in manual and CNN classification experiments. The good expansibility and robustness of the system shown in extending experiments indicate the promising potential for practical applications of SAR target recognition.

The efficiency and accuracy of target recognition in synthetic aperture radar (SAR) imagery have seen significant progress lately, stemming from the encouraging advancements of automatic target recognition (ATR) technology based on deep learning. However, the development of a deep learning-based SAR ATR algorithm still faces two critical challenges: the difficulty of feature extraction caused by the unique nature of SAR imagery and the scarcity of datasets caused by the high acquisition cost. Due to its desirable image nature and extremely low acquisition cost, the simulated optical target imagery obtained through computer simulation is considered a valuable complement to SAR imagery. In this study, a CycleGAN-based SAR and simulated optical image fusion network (SOIF-CycleGAN) is designed and demonstrated to mitigate the adverse effects of both challenges simultaneously through SAR-optical image bidirectional translation. SAR-to-optical (S2O) image translation produces artificial optical images that are high-quality and rich in details, which are used as supplementary information for SAR images to assist ATR. Conversely, optical-to-SAR (O2S) image translation generates pattern-rich artificial SAR images and provides additional training data for SAR ATR algorithms. Meanwhile, a new dataset of SAR-optical image pairs containing eight different types of aircraft has been created for training and testing SOIF-CycleGAN. By combining image-quality assessment (IQA) methods and human vision, the evaluation verified that the proposed network possesses exceptional bidirectional translation capability. Finally, the results of the S2O and O2S image translations are simultaneously integrated into a SAR ATR network, resulting in an overall accuracy improvement of 6.33%. This demonstrates the effectiveness of SAR-optical image fusion in enhancing the performance of SAR ATR.

A novel semantic-aided feature generative adversarial network (SFGAN) is proposed in this letter and utilized in zero-shot learning (ZSL) for fast synthetic aperture radar (SAR) automatic target recognition (ATR). By exploiting shared semantic attributes of all classes, the SFGAN is capable of directly generating and recognizing samples with sufficient discriminability and reality from feature dimension, and ultimately achieves accurate and fast recognition of unseen targets. To enhance the quality and stability of feature space, the Wasserstein distance measurement based on the semantic attributes is applied in generative adversarial network (GAN). Meanwhile, in order to improve the discriminability of generated features, a classification sub-net is integrated into the network. Additionally, Gaussian mixture model (GMM) and reconstruction loss are introduced in feature generation to align the unseen features more consistent with distribution of real data. In experiments on a new self-built SAR aircraft dataset, SFGAN achieves 70.14% accuracy on unseen targets and improves the training speed by more than two times, fully demonstrating the advancement and effectiveness of the proposed method.

In traditional multiband fusion (MF) methods, the differences in the scattering center (SC) among different sub-bands are ignored, which leads to incoherence between sub-band echoes and results in limited estimation accuracy for full-band echo (FBE). To solve this problem, a novel MF method based on differential processing of the SCs in each sub-band is proposed and experimentally demonstrated. In this method, the SCs in each sub-band are classified using a proposed two-layer contrast strategy into two kinds of SCs, which are fused and super-resolved, respectively, to produce the FBE. For application during the classification, fusion, and super-resolution processes, a general iterative shrinkage thresholding algorithm is developed to ensure both the accuracy and speed of estimation of the SC locations and amplitudes. Total FBE is calculated by superimposing the fused FBE and the super-resolved FBE. In experiments, the bandwidth of each sub-band is improved by an average factor of 3.97 when the signal-to-noise ratio is equal to 10 dB and the frequency gap between two sub-bands is equal to 5.08 GHz. Thus, the high estimation accuracy of the proposed method is verified, as well as its high robustness to noise and incoherence.