Min Huang

To address the challenges of high-dimensional state exploration, random wind field disturbances, near-obstacle collision risks, and unstable value estimation in UAV three-dimensional path planning under urban low-altitude complex environments, a path planning algorithm named CL-PER-ID3QN is proposed, which integrates Curriculum Learning (CL), Prioritized Experience Replay (PER), and the Improved Dueling Double Deep Q-Network (ID3QN). Firstly, the ID3QN architecture is proposed, incorporating PReLU and the dynamic learning rate to enhance the model’s feature extraction capability and training stability; the Adaptive-Gated Value-Advantage Aggregation (AG-VAA) mechanism is proposed to improve the ability to discriminate action values. Secondly, a three-dimensional grid environment for random wind-field perception is constructed, and the wind-obstacle coupled risk proxy model is proposed, enabling the UAV to approach the target while accounting for energy consumption and safety. Finally, the CL and PER mechanisms are introduced to alleviate the initial exploration difficulties and increase the utilization rate of key experience samples. Ablation experiments verify the effectiveness of various improvement strategies. Comparative experimental results show that, compared with the baseline D3QN algorithm, the CL-PER-ID3QN algorithm achieves a 10.46 percentage-point increase in Success Rate (SR), reaching 83.59%, and a 4.27 percentage-point decrease in Collision Rate (CR).

To address the problems that existing UAV RF signal azimuth estimation methods rely on idealized simulation data and lack accuracy and robustness in complex environments, a high-precision azimuth estimation method based on an improved ResNet, namely the Dual-Branch Circular-Regression Network, is proposed. Firstly, UCA-8 array data is generated from measured single-channel RF signals, and non-ideal factors such as channel mismatch and mutual coupling among array elements are incorporated to simulate the real RF receiving environment. Secondly, ResNet is improved from three aspects: input normalization, dynamic dual-branch (DDB) learning features and periodic angle regression. The input normalization strategy based on Per-Sample Complex Root Mean Square (PSCRMS) is adopted to improve the adaptability of the model to signal scale changes. The DDB structure is adopted to adaptively fuse I/Q spatiotemporal features with a spatial covariance statistical prior to enhance the spatial feature expression ability in complex scenes. A periodic angle regression method based on Unit Circular Vector Representation (UCVR) and the Huber Loss (GAH Loss) of geodesic angle distance is adopted to realize periodic angle continuous modeling and suppress abnormal angle errors. Finally, comparative and ablation experiments are conducted on the constructed UCA-8 dataset. The experimental results show that compared with the baseline ResNet, the Dual-Branch Circular-Regression Network achieves 85.8%, 95.3%, and 81.5% reductions in MAE, RMSE and P₉₅, respectively, and maintains higher estimation accuracy and good robustness under low signal-to-noise ratio, hardware mismatch and co-frequency dual-source interference.

In the multi-class recognition task for unmanned aerial vehicle (UAV) radio frequency (RF) signals, noise interference and complex backgrounds can significantly degrade recognition performance. The stability and accuracy of existing recognition methods often fail to meet the requirements of practical applications. To enhance the stability and accuracy of UAV RF signal recognition, especially to mitigate performance degradation in complex backgrounds, a UAV RF signal classification method, MD-Net, is proposed that integrates Adaptive Time-Frequency Masking and a dual-network architecture. First, an Adaptive Time-Frequency Masking mechanism is constructed. By analyzing the energy distribution of RF signals in the time-frequency domain, the masking region is automatically determined, ensuring that the training data maintains a diverse distribution across different interference scenarios. This significantly improves the model’s anti-interference performance and discriminative stability in complex environments. Subsequently, a dual-branch recognition network architecture is designed, integrating a multi-layer perceptron (MLP) and a long short-term memory (LSTM) network. The MLP extracts static amplitude features from the signals, while the LSTM learns time-series features. These two feature types are then fused to achieve complementary characteristics, ultimately enabling accurate classification of UAV RF signals. Extensive comparative experiments conducted on the DroneRF dataset demonstrate that the MD-Net model achieves an average recognition accuracy of 85.58%, an improvement of 5.27 percentage points over the baseline model. The experimental results show that Adaptive Time-Frequency Masking can effectively enhance the model’s adaptability to real-world interference environments, while the dual-network fusion mechanism fully integrates static amplitude and time-series features, providing a feasible and highly reliable technical approach for UAV RF signal recognition.

Ray tracing is an important computational method in electromagnetism. To improve the efficiency of electromagnetic simulation, spatial division structures such as the Bounding Volume Hierarchy (BVH) are often used to accelerate calculations. Studies have shown that the traversal efficiency of the BVH tree is also affected to a certain extent by the distribution of scene primitives. In electromagnetic ray tracing with direction sensitivity, the distribution of scene primitives has a greater impact on BVH traversal efficiency. Therefore, to further improve tracing efficiency, this paper proposes a BVH construction method based on the local coherence of triangular primitives and their directional sensitivity. First, this method uses Morton encoding to spatially sort the primitives, thereby enhancing spatial locality within nodes and minimizing the redundant volume of the bounding box. Second, a local adjacency clustering method is proposed to further optimize the data distribution of leaf nodes. Finally, a direction-weighted scoring mechanism is proposed to prioritize traversal of triangular primitives closer to the emission source’s viewing angle and distance, achieving a direction-priority hierarchy and traversal acceleration that conform to the characteristics of electromagnetic wave propagation. Experiments show that in four typical scenarios, such as classrooms, residential areas, schools, and mountainous regions, the calculation results of the improved algorithm, while maintaining consistency with the baseline model, increase the calculation efficiency by an average of ∼15% and accelerate the electromagnetic simulation calculation process.

Synthetic Aperture Radar (SAR), with its outstanding features, has developed into a vital technical method for the global detection of maritime and terrestrial targets. However, the unique imaging mechanism intrinsic to SAR inherently produces coherent speckle noise. This noise significantly deteriorates the quality of the image, thereby adversely affecting downstream applications like target recognition and classification. Therefore, research on SAR image denoising holds significant practical and theoretical value. However, due to the limitations of the local receptive field of CNN, it is difficult for it to capture global spatial features. Although Transformer can achieve global spatial modeling, its quadratic complexity results in a large amount of computational overhead. To tackle these issues, this paper introduces RSS-Net, a new denoising network for SAR images, built on an encoder-decoder architecture. RSS-Net demonstrates significant improvements in SAR image denoising performance, due to its ability to extract multi-scale feature information. The network incorporates the Residue State-Space Block (RSSB), which fuses Mamba's Vision State Space Module (VSSM) and a CNN-based Channel Attention Block (CAB). VSSM leverages 2D-Selective Scan (SS2D) better acquires spatial information features in SAR images by scanning in four directions. RSSB efficiently merges Mamba's ability to model long-range dependencies and conventional convolution's strengths in extracting local features. This integration effectively alleviates the common detail blurring problem in existing SAR denoising methods and strengthens the restoration of high-frequency image details. The novel application of the Mamba to SAR image denoising enables the proposed method to attain both long-range contextual dependency modeling and linear computational cost, thus resolving the balance between global modeling capacity and efficiency in computation, while also pointing to new avenues for future research. Experimental evaluations confirm the superior denoising efficacy of the proposed algorithm.

Synthetic Aperture Radar (SAR) is extensively utilized in diverse fields, including military defense and resource exploration, due to its all-day, all-weather characteristics. However, the extraction of information from SAR images is severely affected by speckle noise, making denoising crucial. This paper proposes a Hybrid Transformer-CNN Network (HTCNet), a hybrid denoising network that combines Transformer and CNN. The three core designs of the network ensure its suitability for SAR image denoising: (1) The network integrates a Transformer-based encoder with a CNN-based decoder, capturing both local and global dependencies inherent in SAR images, thereby enhancing the effectiveness of noise removal. (2) Patch embedding blocks enhance the convolutional neural network's perception of features at different scales. (3) Depthwise separable convolutions are fused into the Transformer block to further improve the network's ability to capture spatial information while reducing computational complexity. The proposed algorithm demonstrates excellent denoising performance in both simulated and real SAR images, as evidenced by experimental results. Compared to other denoising algorithms, this method efficiently removes speckle noise while preserving the texture information within the images. Code link: https://github.com/wl14x/Denoising

In the rapidly developing drone industry, drone use has led to a series of safety hazards in both civil and military settings, making drone detection an increasingly important research field. It is difficult to overcome this challenge with traditional object detection solutions. Based on YOLOv8, we present a lightweight, real-time, and accurate anti-drone detection model (EDGS-YOLOv8). This is performed by improving the model structure, introducing ghost convolution in the neck to reduce the model size, adding efficient multi-scale attention (EMA), and improving the detection head using DCNv2 (deformable convolutional net v2). The proposed method is evaluated using two UAV image datasets, DUT Anti-UAV and Det-Fly, with a comparison to the YOLOv8 baseline model. The results demonstrate that on the DUT Anti-UAV dataset, EDGS-YOLOv8 achieves an AP value of 0.971, which is 3.1% higher than YOLOv8n’s mAP, while maintaining a model size of only 4.23 MB. The research findings and methods outlined here are crucial for improving target detection accuracy and developing lightweight UAV models.

Image denoising, as an essential component of image pre-processing, effectively reduces noise interference to enhance image quality, a factor of considerable research importance. Traditional denoising methods often lead to the blurring of image details and a lack of realism at the image edges. To deal with these issues, we propose an image denoising algorithm named Residual structure and Cooperative Attention mechanism based on Generative Adversarial Networks (RCA-GAN). This algorithm proficiently reduces noise while focusing on preserving image texture details. To maximize feature extraction, this model first employs residual learning within a portion of the generator’s backbone, conducting extensive multi-dimensional feature extraction to preserve a greater amount of image details. Secondly, it introduces a simple yet efficient cooperative attention module to enhance the representation capacity of edge and texture features, further enhancing the preservation of intricate image details. Finally, this paper constructs a novel loss function—the Multimodal Loss Function—for the network training process. The experimental results were evaluated using Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity (SSIM) as evaluation metrics. The experimental results demonstrate that the proposed RCA-GAN image denoising algorithm has increased the average PSNR from 24.71 dB to 33.76 dB, achieving a 36.6% improvement. Additionally, the average SSIM value has risen from 0.8451 to 0.9503, indicating a 12.4% enhancement. It achieves superior visual outcomes, showcasing the ability to preserve image texture details to a greater extent and excel in edge preservation and noise suppression.

Due to the diversity of aircraft target scale and interference of background strong scattering in synthetic aperture radar (SAR) images, it is a challenge for target detection tasks. In response to these problems, this paper proposes a new SAR image aircraft target detection model named EST-YOLOv5s. The proposed model integrates the Efficient Channel Attention (ECA) mechanism into the C3 module of the backbone network, which enhances the scattering features of aircraft targets and suppresses irrelevant background information without increasing the number of parameters. Secondly, replace the bottleneck module in the last C3 module in the backbone network with the Swin Transformer Block. By using the shifted window partitioning approach to obtain the global perception ability, the problem of missed detection of small objects is improved. Finally, the Task-Specific Context Decoupling (TSCODE) head is used to balance the relationship between classification and regression so that different contextual details can be better utilized. In this paper, the SAR Aircraft Detection Dataset (SADD) is used as the experimental data set to compare with the baseline model YOLOv5s. The experimental outcomes indicate that the recall of the EST-YOLOv5s model reached 94.2%, the precision reached 97.3%, and the mAP@50 reached 97.8%, which were 2.3%, 1.7%, and 1.7% higher than YOLOv5s respectively. Furthermore, our model also meets the real-time requirements in terms of speed and exhibits strong anti-interference ability.

Detecting small ship targets in large-scale synthetic aperture radar (SAR) images with complex backgrounds is challenging. This difficulty arises due to indistinct visual features and noise interference. To address these issues, we propose a novel two-stage detector, namely a convolutional and visual transformer fusion network (CViTF-Net), and enhance its detection performance through three innovative modules. Firstly, we designed a pyramid structured CViT backbone. This design leverages convolutional blocks to extract low-level and local features, while utilizing transformer blocks to capture inter-object dependencies over larger image regions. As a result, the CViT backbone adeptly integrates local and global information to bolster the feature representation capacity of targets. Subsequently, we proposed the Gaussian prior discrepancy (GPD) assigner. This assigner employs the discrepancy of Gaussian distributions in two dimensions to assess the degree of matching between priors and ground truth values, thus refining the discriminative criteria for positive and negative samples. Lastly, we designed the level synchronized attention mechanism (LSAM). This mechanism simultaneously considers information from multiple layers in region of interest (RoI) feature maps, and adaptively adjusts the weights of diverse regions within the final RoI. As a result, it enhances the capability to capture both target details and contextual information. We achieved the highest comprehensive evaluation results for the public LS-SSDD-v1.0 dataset, with an mAP of 79.7% and an F1 of 80.8%. In addition, the robustness of the CViTF-Net was validated using the public SSDD dataset. Visualization of the experimental results indicated that CViTF-Net can effectively enhance the detection performance for small ship targets in complex scenes.