Hyperspectral Anomaly Detection via Dual Dictionaries Construction Guided by Two-Stage Complementary Decision

Automatic Modulation Classification Based on CNN-Transformer Graph Neural Network

In recent years, neural network algorithms have demonstrated tremendous potential for modulation classification. Deep learning methods typically take raw signals or convert signals into time-frequency images as inputs to convolutional neural networks (CNNs) or recurrent neural networks (RNNs). However, with the advancement of graph neural networks (GNNs), a new approach has been introduced involving transforming time series data into graph structures. In this study, we propose a CNN-transformer graph neural network (CTGNet) for modulation classification, to uncover complex representations in signal data. First, we apply sliding window processing to the original signals, obtaining signal subsequences and reorganizing them into a signal subsequence matrix. Subsequently, we employ CTGNet, which adaptively maps the preprocessed signal matrices into graph structures, and utilize a graph neural network based on GraphSAGE and DMoNPool for classification. Extensive experiments demonstrated that our method outperformed advanced deep learning techniques, achieving the highest recognition accuracy. This underscores CTGNet's significant advantage in capturing key features in signal data and providing an effective solution for modulation classification tasks.

Co-Design Dedicated System for Efficient Object Tracking Using Swarm Intelligence-Oriented Search Strategies

The template matching technique is one of the most applied methods to find patterns in images, in which a reduced-size image, called a target, is searched within another image that represents the overall environment. In this work, template matching is used via a co-design system. A hardware coprocessor is designed for the computationally demanding step of template matching, which is the calculation of the normalized cross-correlation coefficient. This computation allows invariance in the global brightness changes in the images, but it is computationally more expensive when using images of larger dimensions, or even sets of images. Furthermore, we investigate the performance of six different swarm intelligence techniques aiming to accelerate the target search process. To evaluate the proposed design, the processing time, the number of iterations, and the success rate were compared. The results show that it is possible to obtain approaches capable of processing video images at 30 frames per second with an acceptable average success rate for detecting the tracked target. The search strategies based on PSO, ABC, FFA, and CS are able to meet the processing time of 30 frame/s, yielding average accuracy rates above 80% for the pipelined co-design implementation. However, FWA, EHO, and BFOA could not achieve the required timing restriction, and they achieved an acceptance rate around 60%. Among all the investigated search strategies, the PSO provides the best performance, yielding an average processing time of 16.22 ms coupled with a 95% success rate.

Ship Detection in SAR Images Based on Feature Enhancement Swin Transformer and Adjacent Feature Fusion

Convolutional neural networks (CNNs) have achieved milestones in object detection of synthetic aperture radar (SAR) images. Recently, vision transformers and their variants have shown great promise in detection tasks. However, ship detection in SAR images remains a substantial challenge because of the characteristics of strong scattering, multi-scale, and complex backgrounds of ship objects in SAR images. This paper proposes an enhancement Swin transformer detection network, named ESTDNet, to complete the ship detection in SAR images to solve the above problems. We adopt the Swin transformer of Cascade-R-CNN (Cascade R-CNN Swin) as a benchmark model in ESTDNet. Based on this, we built two modules in ESTDNet: the feature enhancement Swin transformer (FESwin) module for improving feature extraction capability and the adjacent feature fusion (AFF) module for optimizing feature pyramids. Firstly, the FESwin module is employed as the backbone network, aggregating contextual information about perceptions before and after the Swin transformer model using CNN. It uses single-point channel information interaction as the primary and local spatial information interaction as the secondary for scale fusion based on capturing visual dependence through self-attention, which improves spatial-to-channel feature expression and increases the utilization of ship information from SAR images. Secondly, the AFF module is a weighted selection fusion of each high-level feature in the feature pyramid with its adjacent shallow-level features using learnable adaptive weights, allowing the ship information of SAR images to be focused on the feature maps at more scales and improving the recognition and localization capability for ships in SAR images. Finally, the ablation study conducted on the SSDD dataset validates the effectiveness of the two components proposed in the ESTDNet detector. Moreover, the experiments executed on two public datasets consisting of SSDD and SARShip demonstrate that the ESTDNet detector outperforms the state-of-the-art methods, which provides a new idea for ship detection in SAR images.