Telecommunications System & Management

The growing demand for satellite services and the increasing reliance on satellite networks for communication, navigation, and Earth observation have highlighted the importance of sustainability in space-based operations. As more businesses and organizations look to integrate Low Earth Orbit (LEO) satellite networks into their operations, there is an urgent need to address the environmental impact of these systems. Green computing, which focuses on reducing the energy consumption and environmental footprint of technology systems, offers a promising strategy for promoting sustainability in LEO satellite networks. This report explores the role of green computing in business integration within LEO satellite networks, focusing on strategies that can optimize performance while minimizing energy consumption and environmental harm. LEO satellite networks are being deployed to meet the increasing demand for high-speed internet, remote sensing, and global communications. These networks involve constellations of satellites orbiting at altitudes between 300 and 2,000 kilometers, providing near-global coverage and low-latency services.

The advent of networking, computing, and immersive technologies has revolutionized the concept of smart environments, driving advancements across industries such as healthcare, education, entertainment, and urban development. A smart environment is one where physical spaces are integrated with digital technologies to create intelligent, interactive, and responsive spaces. The combination of high-speed communication networks, powerful computing capabilities, and immersive technologies such as Augmented Reality (AR), Virtual Reality (VR), and Mixed Reality (MR) plays a crucial role in enabling these environments. By seamlessly blending the physical and digital worlds, these technologies enhance user experiences, optimize operations, and create opportunities for innovation in various fields. Networking is one of the foundational pillars of smart environments, enabling real-time communication and the exchange of data between devices and systems. In a smart environment, multiple devices, sensors, and actuators are connected to form an intricate network that collects and processes data to drive intelligent decision-making.

The detection of defects in industrial products is a crucial task in quality control and manufacturing processes. Among various components produced in industries, rubber rings are widely used in many applications, including automotive, aerospace, and machinery, where they function as seals and gaskets. Given their importance, ensuring that these rubber rings are defect-free is essential for maintaining product reliability and safety. However, manually inspecting rubber rings for defects is a labor-intensive and error-prone process. With the rapid advancement of machine learning, particularly neural networks, there is significant potential to automate this task. A key challenge, however, lies in designing neural network models that are not only accurate but also efficient in terms of computational resources and processing time. This is particularly important in industrial settings, where real-time performance and resource constraints are often critical factors. The optimization of lightweight neural networks for rubber ring defect detection addresses the need for a model that can perform high-accuracy defect detection while minimizing the computational burden.

Underwater image super-resolution is a challenging task that seeks to enhance the quality of images captured in aquatic environments. Images taken underwater often suffer from various distortions such as low contrast, blurring, and loss of details due to the scattering and absorption of light in water. These issues are exacerbated by the lack of natural lighting, depth, and limited visibility, making it difficult to capture high-resolution images in such conditions. The need for high-quality, high-resolution underwater images have grown significantly in fields like marine biology, underwater robotics, environmental monitoring, and even underwater tourism. Enhancing these images for accurate interpretation and analysis is essential, and this is where advanced computational techniques like deep learning come into play. A promising approach to tackle the challenges of underwater image enhancement is the use of dynamic structure-aware networks, which are designed to handle the inherent complexities of underwater image characteristics and improve resolution.

Domain Generation Algorithms (DGAs) are often used by cybercriminals to create large numbers of domain names that can be used for malicious purposes, such as hosting phishing sites, controlling botnets, or spreading malware. These domains are often difficult to detect because they change frequently, making traditional detection methods ineffective. Therefore, the need for more sophisticated detection techniques has arisen, especially in the context of Domain Name System (DNS) traffic analysis. One promising approach to detecting DGA-generated domains is the application of advanced machine learning models, such as the Transformer and Rapid Selective Kernel Network (RSKN). These methods can significantly improve the accuracy and efficiency of DGA domain detection by leveraging their powerful feature extraction and representation capabilities. The Transformer model, originally designed for Natural Language Processing (NLP) tasks, has shown impressive performance across various domains due to its ability to capture long-range dependencies and learn contextual relationships in data. In the context of DGA domain detection, the Transformer model can be applied to analyze sequences of characters within domain names, which may exhibit distinct patterns or structures compared to legitimate domain names.

Unmanned Aerial Vehicles (UAVs) have emerged as versatile tools across a wide range of applications, including surveillance, search and rescue operations, environmental monitoring, and precision agriculture. The ability to detect small targets reliably and efficiently is critical in these scenarios, where the detection task often involves identifying objects of interest that are small in size relative to the vast field of view captured by the UAV's onboard sensors. However, small target detection presents unique challenges due to factors such as the limited resolution of onboard cameras, varying lighting conditions, complex backgrounds, and the high-speed motion of UAVs. To address these challenges, this report presents a compact algorithm designed specifically for small target detection on UAV platforms. The lightweight nature of UAVs imposes stringent constraints on computational resources and power consumption.

The increasing demands for efficient and reliable performance in aerospace engineering have driven significant advancements in the structural analysis of airfoil components. Airfoils, being critical elements of aircraft and turbine designs, experience complex stress distributions under varying operational conditions. Accurately predicting these stresses is essential to ensure structural integrity, optimize performance, and extend the lifespan of the components. Traditional methods of stress analysis, such as Finite Element Analysis (FEA), are widely used but often involve intensive computational resources and time. To address these challenges, this report explores a Convolutional Neural Network (CNN)-based approach for stress prediction in airfoil structures, offering a novel and efficient solution. Convolutional neural networks have demonstrated remarkable capabilities in extracting spatial features from data, making them particularly suitable for tasks involving image processing and pattern recognition. Leveraging these strengths, the CNN-based method for stress prediction transforms the stress analysis problem into a data-driven learning task.

Ocean eddies play a critical role in marine ecosystems and global climate systems by influencing nutrient distribution, heat transport, and oceanic circulation patterns. Detecting and analyzing these dynamic structures is essential for understanding their impact on ocean processes and for applications such as climate modeling, marine resource management, and navigation. Traditional methods of detecting ocean eddies often rely on satellite altimetry data combined with algorithms that analyze sea surface height anomalies. While these techniques have been effective, they can be computationally intensive and may struggle with real-time detection or identifying smaller eddies. To address these challenges, this report presents an efficient and lightweight convolutional network designed specifically for ocean eddy detection, offering an innovative approach that balances accuracy and computational efficiency. The proposed convolutional network leverages the power of deep learning to process spatial and temporal oceanographic data, identifying eddies with high precision.

Emotion recognition from EEG signals is an area of increasing interest due to its potential applications in healthcare, human-computer interaction, and mental health monitoring. Electroencephalography signals reflect the brain’s electrical activity and offer valuable insights into emotional states. Recognizing emotions from EEG signals, however, is a challenging task due to the complexity and variability of brain activity. Traditional emotion recognition techniques often struggle to capture the intricate patterns in EEG data that are associated with different emotional states. The development of more sophisticated models, such as those based on attention mechanisms, holds promise in improving the accuracy and robustness of emotion recognition systems. This report explores the application of a dual attention mechanism to enhance the performance of EEG signal emotion recognition models. The dual attention mechanism in the context of EEG signal emotion recognition involves the use of two distinct attention mechanisms to focus on relevant features in the EEG data. Attention mechanisms, which are inspired by human cognition, allow models to prioritize important input features and suppress irrelevant ones. In the case of EEG signals, attention mechanisms can help the model focus on specific temporal and spatial patterns in the data that are most indicative of emotional states.

Single-Object Tracking (SOT) is a fundamental problem in computer vision, where the goal is to track the movement of a single object across a sequence of frames in a video. This problem is challenging due to various factors such as object appearance changes, occlusions, and varying lighting conditions. The task becomes even more difficult when tracking objects in real-time or on resource-constrained devices, where computational efficiency is a critical concern. In recent years, Siamese networks have emerged as a powerful framework for SOT due to their ability to learn discriminative features from pairs of images. However, while Siamese networks have shown great promise, challenges remain in improving their efficiency and tracking performance, especially in terms of handling large amounts of data and maintaining robustness in dynamic environments. The proposed approach, a lightweight Siamese network with global correlation for single-object tracking, seeks to address these challenges by focusing on computational efficiency and performance. The key to the success of this approach lies in combining the strengths of Siamese networks with a global correlation mechanism, which enhances the ability to capture long-range dependencies between objects in different frames. By leveraging a lightweight network architecture and global correlation, the model can achieve high tracking accuracy while minimizing computational cost, making it suitable for real-time applications and deployment on devices with limited resources.