The rapid development of remote sensing technology has led to a sharp increase in the amount of synthetic aperture radar (SAR) measurements, which put forward higher requirements for remote sensing image processing. As an important application of SAR, fast and accurate ship detection has always been a research hotspot. In this article, an improved lightweight RetinaNet for ship detection in SAR images is proposed. Compared with the standard RetinaNet, the shallow convolutional layers of the backbone are replaced by ghost modules and the number of the deep convolutional layers is reduced. The spatial and channel attention modules are embedded into the model to enhance detectability. K -means clustering algorithm is applied to adjust the initial aspect ratios of the model. The effectiveness and robustness of the proposed method is demonstrated by numerical experiments with SSDD dataset, Gaofen-3 mini dataset, and a large-scale SAR image of Hisea-1 satellite, it is shown that the proposed method can significantly reduce the floating-point operations and the number of parameters without decreasing the detection accuracy and recall ratio. Moreover, the experimental results also show the proposed model's robustness and the ability to detect ship targets in small datasets.

Ocean surveillance via high-resolution Synthetic Aperture Radar (SAR) imageries has been a hot issue because SAR is able to work in all-day and all-weather conditions. The launch of Chinese Gaofen-3 (GF-3) satellite has provided a large number of SAR imageries, making it possible to marine targets monitoring. However, it is difficult for traditional methods to extract effective features to classify and detect different types of marine targets in SAR images. This paper proposes a convolutional neutral network (CNN) model for marine target classification at patch level and an overall scheme for marine target detection in large-scale SAR images. First, eight types of marine targets in GF-3 SAR images are labelled based on feature analysis, building the datasets for further experiments. As for the classification task at patch level, a novel CNN model with six convolutional layers, three pooling layers, and two fully connected layers has been designed. With respect to the detection part, a Single Shot Multi-box Detector with a multi-resolution input (MR-SSD) is developed, which can extract more features at different resolution versions. In order to detect different targets in large-scale SAR images, a whole workflow including sea-land segmentation, cropping with overlapping, detection with MR-SSD model, coordinates mapping, and predicted boxes consolidation is developed. Experiments based on the GF-3 dataset demonstrate the merits of the proposed methods for marine target classification and detection.

Recently, several magnetic tunnel junction (MTJ)-based non-volatile ternary content-addressable memory (NV-TCAM) cells have been proposed to realize zero standby power. However, they still suffer from low reliability and high power consumption during search operations. To address these issues, we propose a novel MTJ-based NV-TCAM cell, which is composed of 15 transistors and 4 MTJs (15T-4MTJ). By utilizing the differential MTJs with complementary states and positive feedback of cross-coupled inverters for sensing, the proposed 15T-4MTJ NV-TCAM cell can significantly improve the search reliability. Moreover, owing to that there is no static current during search operations, only dynamic charging and discharging current, it can achieve ultra-low power consumption. In addition, by using only one transistor as the critical path between the match-line and GND, its switch delay can be shortened, thereby realizing high-speed search operations. Hybrid CMOS/MTJ simulation results of the 144-bit word circuit show that the proposed 15T-4MTJ NV-TCAM cell can obtain a smaller search error rate of 2.7%, a higher search speed of 0.17 ns, and a lower search energy of 0.17 fJ/bit/search in comparison with other MTJ-based NV-TCAM cells. On the other hand, its write energy of 1.589 pJ/bit is about 3.64x smaller than that of the previously proposed 10T-4MTJ NV-TCAM cell.

Polarimetric synthetic aperture radar (PolSAR) image semantic segmentation is currently of great importance for synthetic aperture radar (SAR) image interpretation, especially in agricultural applications. Several convolutional neural networks (CNNs) have been implemented for SAR image semantic segmentation in urban or land cover applications. However, existing CNNs often break one semantic area into several pieces or confuse the adjacent semantic area in a PolSAR image. To address these issues uniformly, a refined pyramid scene parsing network (PSPNet) is proposed for PolSAR image semantic segmentation in an agricultural areas. Compared to conventional PSPNet architecture, the refined PSPNet adopts a multilevel feature fusion design in its decoder to effectively exploit the features learned from its different encoder branches. Besides, a polarimetric channel attention module is incorporated into the network to capture rich polarimetric features in a PolSAR image. Furthermore, an edge-aware loss function is devised to guide the network to refine pixel-level edge information directly from semantic segmentation prediction, separating easily confused agriculture area with sharp contours. Experimental results on one airborne millimeter-wave PolSAR dataset verify that the proposed network achieves promising semantic segmentation accuracy and preferable spatial consistency.

Radio frequency interference (RFI) is known to jam synthetic aperture radar (SAR) measurements, severely degrading the SAR imaging quality. The suppression of RFI in SAR echo signals is usually an underdetermined blind source separation problem. In this paper, we propose a novel method for multiclass RFI detection and suppression based on the single shot multibox detector (SSD). First, an echo-interference dataset is established by randomly combining the target signal with various types of RFI in a simulation, and the time–frequency form of the dataset is obtained by utilizing the short-time Fourier transform (STFT). Next, the time–frequency dataset acts as input data to train the SSD and obtain a network that is capable of detecting, identifying and estimating the interference. Finally, all of the interference signals are exactly reconstructed based on the prediction results of the SSD and mitigated by an adaptive filter. The proposed method can effectively increase the signal-to-interference-noise ratio (SINR) of RFI-contaminated SAR echoes and improve the peak sidelobe ratio (PSLR) after pulse compression. The simulated experimental results validate the effectiveness of the proposed method.

An autofocus operation is an indispensable procedure to obtain well-focused images for synthetic-aperture radar (SAR) systems without precise navigation devices. Three challenges have been faced in the very high-resolution (VHR) airborne SAR autofocus due to the long cumulative time: the varying along-track velocity, the residual range cell migration (RCM), and the range-dependent phase errors with higher order components. When it comes to the stripmap mode, the autofocus becomes more complicated, since the scenario with a few strong scatterers is more likely to be encountered with a moving beam. Combining the merits of parametric and nonparametric autofocus algorithms, a robust motion error estimation method is proposed in this paper. First, we perform a stripmap multiaperture mapdrift autofocus operation to extract the along-track velocity and the most range-invariant errors, removing the residual RCM at a subaperture scale. Second, one referential center block is selected to retrieve the residual range-invariant error, which can eliminate the residual RCM globally in the range dimension. With a global high-quality input, the residual range-variant phase errors can be retrieved precisely utilizing a center-to-edge local maximum-likelihood weighted phase gradient autofocus kernel at last. Experiments on real VHR airborne stripmap SAR data are performed to demonstrate the robustness of the proposed method.