Activation functions are pivotal in complex-valued convolutional neural networks (CV-CNN) as they govern the nonlinear learning of amplitude-phase relationships in polarimetric synthetic aperture radar (PolSAR) data. This letter proposes a Swish-Cardioid activation function (SCAF), which introduces a adaptive gating mechanism inspired by Swish’s adaptive properties into the Cardioid activation framework. SCAF multiplies the activation value by the output of a gating function, enabling automatic adjustment of nonlinear intensity and output range based on input characteristics. This innovation enhances the model’s flexibility and adaptability in handling complex electromagnetic scattering patterns. Experimental results demonstrate that the proposed method achieves superior performance in terrain classification tasks.

The radar echo of a target can be modeled as a 2-D signal whose variables are the intra-pulse sampling time (fast-time) and the inter-pulse sampling time (slow-time). The fast-time instantaneous frequency (FIF) and slow-time instantaneous frequency (SIF) of the signal are modulated by the target's slant range. When the target undergoes complex motion, the radar echo becomes a 2-D signal with fast-varying instantaneous frequencies (IFs). IF analysis for such a signal is challenging. To solve this issue, an extending complex-lag time-frequency distribution (ECTD) is introduced. ECTD is a 3-D distribution for 2-D signals based on traditional complex-lag time-frequency distribution (CTD), and it inherits the good performance in handling fast-varying IFs. By introducing complex-lags in both the fast-time and slow-time dimensions, the ECTD can accurately estimate the SIF and FIF of a 2-D signal with fast-varying IFs. Finally, a reduced interference realization of the ECTD achieved by introducing the frequency domain filter is given. Numerical examples validate the effectiveness of the ECTD. The source code is provided athttps://github.com/XinhangZhu/ECTD.

In a synthetic aperture radar (SAR) system, the three-dimensional rotation of ship targets in the presence of medium to high sea states could cause Doppler frequency shifts and image defocusing, and even the defocusing phenomenon is space-variant along the range direction would occur. These issues would adversely affect the subsequent interpretation of ship targets in SAR images. This paper proposes a refocusing method based on a phase error estimation network (PEE-Net) to address the refocusing problem of three-dimensional rotating ship targets. The proposed method transforms the defocused complex-valued SAR ship image into the range-Doppler domain, estimates the phase error by range unit using PEE-Net, and compensates for space-variant phase errors. To train the network in an unsupervised manner and avoid the challenging task of obtaining labeling samples of non-cooperative ships, the proposed method introduces the image entropy loss function based on the minimum entropy criterion.

In recent years, the complex-valued convolutional neural network (CV-CNN) for processing complex data has made great use in the field of SAR data processing. In this paper, a complex-valued activation enhancement method named CRIA is constructed based on the cross-fusion of real and imaginary activation in the activation layer of CV-CNN, the core of which is to cross-combine the real and imaginary parts of the activation output of the two activation functions to enhance the overall processing of complex data, to enhance the ability of the network to parse complex value information. By conducting classification experiments on ship slices in SAR images of complex data, the experimental results show that the CRIA method in the activation layer can accelerate the network convergence speed and enhance the network classification performance.

Synthetic Aperture Radar (SAR) has advantages such as all-weather, high-resolution, and large mapping bands, which make up for the shortcomings of other surveillance and reconnaissance methods such as optics and infrared. It has good detection and imaging effects on stationary targets. However, SAR imaging of moving targets often results in positional shift and defocusing. The main reason for these problems is that SAR has poor velocity resolution and cannot obtain effective velocity information of moving ship targets. In this case, the paper intends to use High-Frequency Ground Wave Radar (HFSWR) with high velocity resolution to assist SAR [1]. Using the velocity information extracted from HFSWR to achieve position compensation and refocusing of moving ship targets in SAR.

Change detection (CD) methods for remote sensing images based on deep learning have garnered increasing research attention. However, existing deep learning approaches are often tailored for specific types of sensors. Extending these methods to dual-sensor scenarios presents challenges, including difficulties in data fusion and an increase in parameter numbers. To address these challenges, we propose a novel dual-stream encoder–decoder CD network architecture. In the encoder, the architecture comprises a shared-weight Siamese Unet stream for each sensor, with unique weights for different sensors. Before the decoder, a 3-D attention module (3-D AM) is incorporated, processing encoder outputs and fusing features from different streams. In addition, to mitigate the increased model parameter numbers due to the use of dual sensors, we propose a lightweight Unet architecture along with a time-difference structure in each stream. The proposed model is evaluated across multiple scenarios on a dual-sensor CD dataset, yielding an F1 score of 0.572 and the parameter number of 0.91 M. These results showcase high performance on a cost-effective level. Our code is available athttps://github.com/CodeofHuang/SmaDS_SiamUnet.

Due to the large imaging scene of geosynchronous synthetic aperture radar (GEO SAR), multiple ship targets are more likely to appear in the same imaging scene. When the targets overlap in the range dimension, their echo signals after range compression will also overlap and interfere with each other. Besides, the stop-and-go assumption is no longer valid for GEO SAR, and a long coherent processing interval (CPI) will result in significant range migration. In such a scenario, essential imaging processes, including the signal division algorithm for eliminating the overlap interference and the focusing process for echo signal parameter estimation, are difficult to implement. To address these issues, a signal division and focusing algorithm based on multiscale fast-time delay fractional Fourier transform (MFDFrFT) is proposed in this article. To describe the relationship between slant range and propagation distance under the non-stop-and-go assumption, a conversion ratio is introduced into the slant range model. Then, the echo signal model can be characterized as a multicomponent 2-D quadratic phase signal with fast-time delay (2-D FD-QPS). In order to directly analyze the parameters of this signal, a linear reversible transformation named MFDFrFT is proposed. Then, the echo signal of each individual target can be divided, and the estimated parameters can be achieved. With the estimated parameters, the focusing process can be implemented to finally obtain the well-focused images of the targets. Simulation and equivalent experiments are provided to validate the effectiveness of the proposed algorithm.

Ship targets have complicated motions such as random swings that change with the waves, which makes the target defocused and azimuth blurred in Synthetic Aperture Radar (SAR) images, and the classification accuracy of three-dimensional rotating ship targets is low. This paper proposes a mixed-type complex-valued convolutional neural network (Mix-CV-CNN). Mix-CV-CNN can make full use of the amplitude and phase information of complex SAR images, and can better complete the classification of SAR three-dimensional rotating ship targets without refocusing the target. Through SAR three-dimensional rotating ship target simulation analysis and actual measurement data verification, the proposed complex-valued convolutional neural network is experimentally analyzed. The superiority of the network improves the accuracy and reliability of SAR three-dimensional rotating ship target classification.

Aiming at the three-dimensional rotation speed estimation task of a synthetic aperture radar (SAR) ship target, this paper proposes a complex-valued convolutional long short-term memory (CV-ConvLSTM) network. It can simultaneously perceive the time, space and frequency domain information of the complex SAR imagery sequence. All elements of convolutional long short-term memory (ConvLSTM) network including convolutional layer, activation function, input gate, forget gate, and output gate are extended to the complex domain. In order to verify the superiority of CV-ConvLSTM in frequency domain information perception over ConvLSTM. Experiments show that CV-ConvLSTM has higher estimation accuracy than ConvLSTM.

Synthetic aperture radar (SAR) system is susceptible to electromagnetic jamming during imaging, which seriously affects the subsequent interpretation of SAR images. Aiming at the problem of active suppressive jamming, this paper proposes a suppression method of SAR suppressive jamming based on self-supervised complex-valued deep learning, which consists of a novel complex-valued jamming suppression network CV-MUNet++ and a self-supervised training strategy. CV-MUNet++ could fully use the amplitude and phase information of complex-valued SAR images. The network’s weights, activation functions, and convolution operations are designed for complex domain processing. The different information representations of target and jamming in amplitude and phase in SAR images are mined to achieve jamming suppression. The self-supervised training strategy is proposed to solve the problem of relying heavily on manually labeled samples in the traditional network training process and is suitable for application scenarios where ground truth is difficult to obtain under complex jamming. The experimental results show that the proposed method could effectively suppress the active jamming of complex backgrounds and has the ability to self-supervised intelligent jamming suppression.

In Synthetic Aperture Radar (SAR) images, ship targets suffer from blurring due to pitch, yaw, and sway, resulting in poor recognition accuracy. This paper proposes a complex-valued convolutional neural network (CV-CNN) architecture called CV-RotNet. This neural network could realize the recognition of SAR defocused ship targets without three-dimensional rotation refocusing. CV-RotNet makes full use of the amplitude and phase information of SAR images. Based on the classic deep learning architecture, RotNet and CV-RotNet were designed from the real domain and the complex domain. CV-RotNet and RotNet are tested on the five types of SAR three-dimensional rotating target simulation samples and the three types of GF-3 real ship target samples. Experimental results show that the average accuracy of CV-RotNet is higher than RotNet with the same degree of freedom, which reflects the advantages of CV-RotNet over RotNet.

In synthetic aperture radar (SAR) images, ship targets are defocused due to three-dimensional rotation, which affects subsequent SAR target detection and recognition tasks. This paper proposes a complex-valued convolutional neural network (CV-CNN) structure called CV-RefocusNet to refocus SAR three-dimensional rotating ship targets. CV-RefocusNet includes two parts of feature extraction network and image reconstruction network and adopts an end-to-end design method. To make full use of the amplitude and phase information of complex SAR images, the convolutional layer, deconvolutional layer, and activation function in CV-RefocusNet are all extended to the complex domain. Then refocusing experiments on simulated SAR data and GF-3 SAR data show that CV-RefocusNet could further improve the focus accuracy instead of real-value CNN (RV-CNN) with the same degree of freedom.

This letter proposes a complex-valued convolutional gated recurrent unit (CV-ConvGRU) network for the three-dimensional rotation refocusing task of a synthetic aperture radar (SAR) ship target. To take advantage of the amplitude and phase information of complex SAR images, all elements of CV-ConvGRU, including the convolutional layer, activation function, update gate and reset gate, are extended to the complex domain. Based on CV-ConvGRU, a complex-valued SAR ship refocusing network (CV-SSRN) architecture is designed for refocusing experiments. To verify the robustness of the proposed CV-ConvGRU over ConvGRU on information perception, this letter also raises a real-valued SAR ship refocusing network (RV-SSRN), which has the same degree of freedom as CV-SSRN. Finally, experiments are carried out, and all results show the superiority of the proposed method on refocusing accuracy.

Strong reflections from the marine surface reduce the contrast between the target-of-interest and the background in synthetic aperture radar (SAR) images and severely affect the interpretation of the image. This letter proposes a framework of SAR sea clutter suppression based on a new self-supervised training strategy referred to as Clutter2Clutter (C2C), which mines self-supervised information from a large number of unlabeled SAR patches for network training. This letter also proposes a complex-valued UNet++ (CV-UNet++) network model to make full use of both amplitude and phase information of the complex SAR image, and the C2C strategy is used to train the CV-UNet++ for sea clutter suppression. Experiments on GF-3 and TerraSAR-X SAR data show that the proposed method has a better effect on suppressing sea clutter and is able to preserve the target-of-interest energy well.

In the synthetic aperture radar (SAR) images, moving ship targets are defocused due to the movement, which leads to the problem of poor classification accuracy. Therefore, this paper proposes an amplitude-phase-type complex-valued convolutional neural network (AP-CV-CNN) architecture called CV-MotionNet to classify SAR moving ship targets without motion compensation. It utilizes both amplitude and phase information of complex SAR images. CV-MotionNet uses amplitude-phase-type activation function to processing amplitude and phase information more conducive. Then, the proposed CV-MotionNet is tested on simulated five-types SAR moving ship target classification task and GF-3 SAR ship classification. Simulation and experiment show that the classification error can be further reduced if using CV-MotionNet instead of real-valued CNN (RV-CNN) with the same degree of freedom.

With the successful application of convolution neural network (CNN) in image recognition field, this paper presents the complex-valued convolutional network (CV-CNN) using different activation functions. Then, four different activation functions of sigmoid, tanh, Leaky-ReLU and ELU were tested in typical polarimetric SAR image classification tasks. Experiments on benchmark datasets of Flevoland shows that CV-CNN using ELU activation function performs the best, with faster convergence speed and much higher recognition rate.

In this paper, a novel moving target detection and tracking approach is presented based on Gaussian mixture probability hypothesis density (GMPHD) filter in synthetic aperture radar (SAR). Based on the advantages of GMPHD filter and the characteristics of moving target in SAR imagery, GMPHD filter is employed on potential moving target candidates extracted from a sequence of temporal and spatial sub-aperture SAR images to detect and track the moving targets in heavy clutter environment. Utilizing the tracking algorithm, the states and the number of moving targets are obtained over time. Simultaneously, the strong ground stationary clutter lacking of the dynamic behavior is eliminated finally. Both simulation and real Gotcha data set processing results are provided to demonstrate the effectiveness of the proposed approach.