Deceptive jamming techniques against Synthetic Aperture Radar (SAR) are crucial for defending against hostile SAR reconnaissance and safeguarding our sensitive targets and regions. The current library of deceptive jamming templates for SAR is severely insufficient. The acquisition of measured and simulated SAR target images, which are key sources for the SAR deceptive jamming template library, can be problematic due to their lengthy acquisition period and high cost. In contrast, emulational optical target images are easier to generate. To address the challenge of quickly translating rich optical target images into SAR target deceptive jamming templates of diverse poses and high fidelity, this paper proposes a SAR target deceptive jamming template generation method based on Optical-to-SAR Cycle-consistent Adversarial Networks (OTSCycGAN). Firstly, this method utilizes deformable convolutional layers to partially replace the vanilla convolutional layers in the generator, enhancing the adaptive extraction ability of geometric transformation features between samples with varying azimuth angles, elevation angles, and target types in the dataset. Furthermore, a convolution kernel attention mechanism is introduced to dynamically adjust the receptive field sizes, increasing sensitivity for various target scales. Finally, we employ a loss function that incorporates Wasserstein GAN-Gradient Penalty (WGAN-GP) loss, hybrid L1-L2 loss, and focal frequency loss to advance the image quality at the pixel level. Our experiments on the constructed Target-Optical/SAR dataset demonstrate that this method brings a significant performance boost in SAR deceptive jamming template generation. The results of the stripmap SAR deceptive jamming, using deceptive jamming templates generated by OTSCycGAN, further confirm their effectiveness.

There has been increased interest in recognizing the dynamic and flexible changes in shipborne multi-function radar (MFR) working modes. The working modes determine the distribution of pulse descriptor words (PDWs). However, building the mapping relationship from PDWs to working modes in reconnaissance systems presents many challenges, such as the duration of the working modes not being fixed, incomplete temporal features in short PDW slices, and delayed feedback of the reconnaissance information in long PDW slices. This paper recommends an MFR working mode recognition method based on the ShakeDrop regularization dual-path attention temporal convolutional network (DP-ATCN) with prolonged temporal feature preservation. The method uses a temporal feature extraction network with the Convolutional Block Attention Module (CBAM) and ShakeDrop regularization to acquire a high-dimensional space mapping of temporal features of the PDWs in a short time slice. Additionally, with prolonged PDW accumulation, an enhanced TCN is introduced to attain the temporal variation of long-term dependence. This way, secondary correction of MFR working mode recognition results is achieved with both promptness and accuracy. Experimental results and analysis confirm that, despite the presence of missing and spurious pulses, the recommended method performs effectively and consistently in shipborne MFR working mode recognition tasks.

The occurrence of dense interference substantially impairs the imaging performance of Sentinel-1 Synthetic Aperture Radar (SAR), leading to a notable degradation in image quality. Existing interference suppression methods are particularly susceptible to signal loss and imaging defocus, as dense interference and target echoes are heavily coupled in multi-dimensional domains. This paper introduces a deep dense interference suppression network based on a multiple attention mechanism for Sentinel-1 SAR data (AMDIS-Net). It reformulates the dense interference suppression challenge as an end-to-end mapping problem. By mapping the highly coupled dense interference and target echoes to a high-dimensional separable feature space, the AMDIS-Net effectively suppress dense interference. The AMDIS-Net leverage multiple attention mechanism to demonstrate the feature dependence of data between the local and global in frequency domain, thereby effectively mitigating dense interference and minimizing the loss of target echoes. The AMDIS-Net is trained on simulated dense interference datasets, then applied to real dense interference SAR data collected by Sentinel-1. Extensive experimental results illustrate the robust performance of the AMDIS-Net in mitigating dense interference.

The primary objective of jamming strategy optimization is to ensure that a jammer timely finds an effective jamming strategy against the multifunction radar (MFR), thereby ensuring the safety of targets. Deep reinforcement learning (DRL) has been widely applied in solving the problem of jamming strategy optimization. However, the process still faces challenges such as low learning efficiency and a heavy memory burden. Therefore, we propose a fast jamming strategy optimization method with imperfect experience. Firstly, we model the radar countermeasure process as a Markov decision process (MDP), and formulate the jamming reward function by combining the jamming effectiveness and the jammer’s operational intent. Secondly, we design a novel hybrid jamming strategy choice module, which uses imperfect experience to improve the optimization efficiency of jamming strategy. Furthermore, to improve sample efficiency and reduce forgetting caused by small replay buffer, we respectively employ a mixed replay buffer strategy and a knowledge consolidation technique. Finally, extensive experiments demonstrate that under the guidance of imperfect experience, our proposed method achieves faster convergence speed and higher strategy accuracy compared with existing DRL-based methods.

As a critical component of radar countermeasure technology, the effectiveness of a jamming strategy directly influences the success of electromagnetic spectrum operations. However, the partial observability of radar information presents significant challenges for jamming strategy generation. Since the jammer can only infer the radar's working state from limited parameter information, uncertainty in strategy input and evaluation errors in jamming effectiveness may arise, ultimately reducing the performance of the generated strategy. To address these challenges, we model the radar countermeasure process as a partially observable Markov decision process (POMDP) and propose a jamming strategy generation method via self-supervised belief representation learning. Specifically, we integrate a belief representation module into the conventional reinforcement learning (RL) framework. This module captures latent representations of belief states with temporal dependencies using a self-supervised loss, thereby providing more reliable inputs for strategy generation. Furthermore, we refine the jamming reward function to alleviate the impact of evaluation errors. The numerical results in diverse partially observable environments show that the proposed method achieves higher cumulative jamming rewards and improved robustness than existing RL-based methods, significantly improving the probability of successful jamming.