Unmanned Aerial Vehicles (UAVs) are indispensable in promoting the development of remote sensing technology. Nevertheless, the tasks of object recognition in remote sensing images based on UAV platforms face major difficulties and challenges due to the complex and variable background environments and the high-density distribution of objects. This paper proposes an object detection algorithm for UAV remote sensing images—YOLO-Remote, which aims to improve detection accuracy by enhancing YOLOv8. This algorithm innovatively integrates the SaElayer module to enhance the focus on remote sensing targets and improve network efficiency. Additionally, it introduces the Efficient-SPPF structure, which effectively expands the network’s receptive field and promotes deep learning capabilities. To address sample imbalance and improve bounding box localization and classification performance, the study also designs the Focaler-MDPIOU strategy. With these comprehensive optimizations, YOLO-Remote achieves significant progress in network architecture. Experiments were conducted on the NWPU VHR10 and RSOD datasets, and the experimental results show that compared to the base model YOLOv8n, the improved model’s average precision increased by 2.7% and 3.2% respectively, demonstrating its superiority in the field of object detection for UAV remote sensing images.The code is available at https://github.com/QuincyQAQ/Yolo-Remote.
4 March 2025 in IEEE Transactions on Consumer Electronics
With the rapid improvement of medical consumer electronics (CE), the transmission of multiple images has become increasingly common in the medical field. During the transmission and storage of image data, sensitive information, including medical images, may be accessed by unauthorized parties, resulting in possible privacy violations. Consequently, several multi-image encryption (MIE) algorithms have been proposed to solve this real-world problem. Nevertheless, current MIE schemes exhibit several drawbacks, including insufficient security in permutation and diffusion algorithms, as well as the insensitivity of chaotic maps. This paper presents a new encryption algorithm called DPPAD-IE. The algorithm primarily consists of three components. First, we developed an innovative hyperchaotic map called the 2D Cross Gaussian Hyperchaotic Map (2D-CGHM), which has an expanded chaotic interval and is proficient in producing chaotic sequences with improved pseudorandomness. Second, we introduce a new Dynamic Polyhedra Permutation algorithm (DPP) that successfully alters pixel locations of the image, hence diminishing the connection between neighboring pixels. Third, we propose an Arnold Diffusion algorithm (AD), using chaotic sequences to further encrypt the permuted image. The DPPAD-IE is capable of encrypting images of any quantity and size concurrently. The experimental results confirm that this approach is both robust and effective, capable of defending against various security attacks. The code are available at:https://github.com/QuincyQAQ/DPPAD-IE.
Graph representation learning aims to preserve graph topology when mapping nodes to vector representations, enabling downstream tasks like node classification and community detection. However, most graph neural network models require extensive labelled data, limiting their practical applicability. To address this, researchers have explored Graph Contrastive Learning (GCL), which uses enhanced graph data and contrastive learning to better capture graph structure and features, providing new avenues for solving real-world problems with limited labelled data. Building on this, this work proposes Graph Representation Embedding Enhanced via Multidimensional Contrastive Learning (GRE2-MDCL). GRE2-MDCL first globally and locally augments the input graph using SVD and LAGNN. The enhanced data is then fed into a triple network with a multi-head attention GNN as the core model. Finally, GRE2-MDCL constructs a multidimensional contrastive loss, incorporating cross-network, cross-view, and neighbor contrast, to optimize the model. Evaluated on Cora, Citeseer, and PubMed, GRE2-MDCL achieves average accuracies of 83.1%, 72.6%, and 82.7%, outperforming baseline GCL models. Visualizations also show tighter intra-cluster aggregation and clearer inter-cluster boundaries, demonstrating the framework’s effectiveness in improving upon the baseline.