A Cross-Stage Focused Small Object Detection Network for Unmanned Aerial Vehicle Assisted Maritime Applications

Journal of Marine Science and Engineering, Journal Year: 2025, Volume and Issue: 13(1), P. 82 - 82

Published: Jan. 5, 2025

The application potential of unmanned aerial vehicles (UAVs) in marine search and rescue is especially concern for the ongoing advancement visual recognition technology image processing technology. Limited computing resources, insufficient pixel representation small objects high-altitude images, challenging visibility conditions hinder UAVs’ target performance maritime operations, highlighting need further optimization enhancement. This study introduces an innovative detection framework, CFSD-UAVNet, designed to boost accuracy detecting minor within imagery captured from elevated altitudes. To improve feature pyramid network (FPN) path aggregation (PAN), a newly PHead structure was proposed, focusing on better leveraging shallow features. Then, structural pruning applied refine model enhance its capability objects. Moreover, conserve computational lightweight CED module introduced reduce parameters resources UAV. At same time, each layer, CRE integrated, attention mechanisms heads precision object detection. Finally, model’s robustness, WIoUv2 loss function employed, ensuring balanced treatment positive negative samples. CFSD-UAVNet evaluated publicly available SeaDronesSee dataset compared with other cutting-edge algorithms. experimental results showed that achieved mAP@50 80.1% only 1.7 M cost 10.2 G, marking 12.1% improvement over YOLOv8 4.6% increase DETR. novel effectively balances limitations scenarios accuracy, demonstrating value field UAV-assisted rescue.

Language: Английский

---

ARTIFICIAL INTELLIGENCE SUPPORT IN DISASTER MANAGEMENT

Kamu Yönetimi ve Teknoloji Dergisi, Journal Year: 2025, Volume and Issue: 7(1), P. 13 - 36

Published: Jan. 30, 2025

The rapid development of digital technologies has driven significant advancements in artificial intelligence (AI) applications, expanding their use across various fields. One notable area is disaster management, where AI leveraged to strengthen societal resilience and protect communities from disasters. However, some projects may fall short expectations during implementation, often resulting increased costs, time, labor due inherent complexity. In response, this study presents a model that explores the application throughout management process, utilizing secondary data sources. objective contribute both academic literature practices by supporting prevention, reducing loss life property, enabling more efficient timely interventions. Furthermore, aims serve as valuable resource not only for researchers field but also decision-makers practitioners, offering concise reference informed, data-driven actions.

Language: Английский

---

Drone selection for disaster responses: an application of the stratified-best-worst method

Management Decision, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 19, 2025

Purpose The purpose of this research is to address the challenges selecting optimal drones for disaster response operations under uncertainties. Traditional static (deterministic) models often fail capture complexities and uncertainties scenarios. This study aims develop a more resilient adaptable decision-making framework by integrating best-worst method (BWM) with stratified multi-criteria (SMCDM), focusing on various uncertainty scenarios such as weather conditions, communication navigation control issues. Design/methodology/approach methodology involves identifying seven essential criteria drone evaluation, guided contingency theory. BWM derives weights each criterion comparing best worst alternatives. SMCDM incorporates different into process. Sensitivity analysis assesses robustness decisions weightings operational integrated approach demonstrated through practical application Kerala flood scenario. Findings proves be highly effective in adapting scenarios, enabling decision-makers consistently identify response. method’s ability account uncertain conditions weather, issues ensures that selected based situation at hand. Research limitations/implications fills critical gaps literature offering comprehensive model selection. However, there are certain limitations. reliance expert opinions introduces subjectivity, potentially affecting generalizability results. In addition, study’s focus single case, floods, limits its applicability other geographic contexts. Integrating real-time data analytics process could also enhance model’s adaptability evolving improve relevance. Practical implications offers practical, By SMCDM, can uncertainties, or disruptions, make informed choices. leads better resource allocation efficient operations, ultimately enhancing speed effectiveness relief efforts adjust scenario-specific factors optimally deployed according unique demands disaster. Social incorporating proposed assists appropriately choosing their characteristics crucial specific thereby efficiency operations. Originality/value presents integration creating dynamic selection addresses posed environments. Unlike traditional methods, allows resulting reliable responsive deployment. bridges gap existing tool response, providing new insights applications optimizing complex, real-world

Language: Английский

---

Edge and Cloud Computing in Smart Cities

Future Internet, Journal Year: 2025, Volume and Issue: 17(3), P. 118 - 118

Published: March 6, 2025

The evolution of smart cities is intrinsically linked to advancements in computing paradigms that support real-time data processing, intelligent decision-making, and efficient resource utilization. Edge cloud have emerged as fundamental pillars enable scalable, distributed, latency-aware services urban environments. Cloud provides extensive computational capabilities centralized storage, whereas edge ensures localized processing mitigate network congestion latency. This survey presents an in-depth analysis the integration cities, highlighting architectural frameworks, enabling technologies, application domains, key research challenges. study examines allocation strategies, analytics, security considerations, emphasizing synergies trade-offs between paradigms. present also notes future directions address critical challenges, paving way for sustainable development.

Language: Английский

---

Real-Time Object Detection Based on UAV Remote Sensing: A Systematic Literature Review

Drones, Journal Year: 2023, Volume and Issue: 7(10), P. 620 - 620

Published: Oct. 3, 2023

Real-time object detection based on UAV remote sensing is widely required in different scenarios. In the past 20 years, with development of unmanned aerial vehicles (UAV), technology, deep learning and edge computing research real-time fields has become increasingly important. However, since a comprehensive task involving hardware, algorithms, other components, complete implementation often overlooked. Although there large amount literature sensing, little attention been given to its workflow. This paper aims systematically review previous studies about from application scenarios, hardware selection, paradigms, algorithms their optimization technologies, evaluation metrics. Through visual narrative analyses, conclusions cover all proposed questions. more demand scenarios such as emergency rescue precision agriculture. Multi-rotor UAVs RGB images are interest applications, mainly uses documented processing strategies. GPU-based platforms used, preferred for detection. Meanwhile, need be focused resource-limited platform deployment, lightweight convolutional layers, etc. addition accuracy, speed, latency, energy equally important Finally, this thoroughly discusses challenges sensor-, computing-, algorithm-related technologies It also prospective impact future developments autonomous communications target

Language: Английский

---