A comprehensive survey of research towards AI-enabled unmanned aerial systems in pre-, active-, and post-wildfire management

Information Fusion, Journal Year: 2024, Volume and Issue: 108, P. 102369 - 102369

Published: March 22, 2024

Wildfires have emerged as one of the most destructive natural disasters worldwide, causing catastrophic losses. These losses underscored urgent need to improve public knowledge and advance existing techniques in wildfire management. Recently, use Artificial Intelligence (AI) wildfires, propelled by integration Unmanned Aerial Vehicles (UAVs) deep learning models, has created an unprecedented momentum implement develop more effective Although survey papers explored learning-based approaches wildfire, drone disaster management, risk assessment, a comprehensive review emphasizing application AI-enabled UAV systems investigating role methods throughout overall workflow multi-stage including pre-fire (e.g., vision-based vegetation fuel measurement), active-fire fire growth modeling), post-fire tasks evacuation planning) is notably lacking. This synthesizes integrates state-of-the-science reviews research at nexus observations modeling, AI, UAVs - topics forefront advances elucidating AI performing monitoring actuation from pre-fire, through stage, To this aim, we provide extensive analysis remote sensing with particular focus on advancements, device specifications, sensor technologies relevant We also examine management approaches, monitoring, prevention strategies, well planning, damage operation strategies. Additionally, summarize wide range computer vision emphasis Machine Learning (ML), Reinforcement (RL), Deep (DL) algorithms for classification, segmentation, detection, tasks. Ultimately, underscore substantial advancement modeling cutting-edge UAV-based data, providing novel insights enhanced predictive capabilities understand dynamic behavior.

Language: Английский

---

Land-Cover Classification Using Deep Learning with High-Resolution Remote-Sensing Imagery

Applied Sciences, Journal Year: 2024, Volume and Issue: 14(5), P. 1844 - 1844

Published: Feb. 23, 2024

Land-area classification (LAC) research offers a promising avenue to address the intricacies of urban planning, agricultural zoning, and environmental monitoring, with specific focus on areas their complex land usage patterns. The potential LAC is significantly propelled by advancements in high-resolution satellite imagery machine learning strategies, particularly use convolutional neural networks (CNNs). Accurate paramount for informed development effective management. Traditional remote-sensing methods encounter limitations precisely classifying dynamic areas. Therefore, this study, we investigated application transfer Inception-v3 DenseNet121 architectures establish reliable system identifying classes. Leveraging these models provided distinct advantages, as it allows benefit from pre-trained features large datasets, enhancing model generalization performance compared starting scratch. Transfer also facilitates utilization limited labeled data fine-tuning, making valuable strategy optimizing accuracy tasks. Moreover, strategically employ fine-tuned versions networks, emphasizing transformative impact architectures. fine-tuning process enables leverage pre-existing knowledge extensive its adaptability LC classification. By aligning advanced techniques, our not only contributes evolution methodologies but underscores importance incorporating cutting-edge methodologies, such network architectures, continual enhancement systems. Through experiments conducted UC-Merced_LandUse dataset, demonstrate effectiveness approach, achieving remarkable results, including 92% accuracy, 93% recall, precision, F1-score. employing heatmap analysis further elucidates decision-making models, providing insights into mechanism. successful CNNs LAC, coupled analysis, opens avenues enhanced monitoring through more accurate automated land-area

Language: Английский

---

Machine learning for polyphenol-based materials

Smart Materials in Medicine, Journal Year: 2024, Volume and Issue: 5(2), P. 221 - 239

Published: Feb. 11, 2024

Language: Английский

---

Experimental Nonlinear and Incremental Control Stabilization of a Tail-Sitter UAV with Hardware-in-the-Loop Validation

Published: Feb. 8, 2024

Tail-sitters aim to combine the advantages of fixed-wing and rotor-craft, but demand a robust fast stabilization strategy perform vertical maneuvers transitions from aerodynamic flight. The research conducted in this work intends assess performance nonlinear control strategies stabilize attitude X-Vert VTOL aircraft when hovering, comparing existing solutions applications Nonlinear Dynamics Inversion (NDI) its incremental version, INDI. Such controllers are implemented tuned simulation order model tail-sitter , complemented by estimation methods that allow feed back necessary variables. These estimators then microcontroller, validating them Hardware-in-the-Loop (HITL) scenario, with simple Lastly, developed used experimental flight, being monitored motion capture system. results validate compare effectiveness different stabilizing it, INDI presenting itself as more strategy, better tracking capabilities less actuator demands.

Language: Английский

---

Precise extraction of targeted apple tree canopy with YOLO-Fi model for advanced UAV spraying plans

Computers and Electronics in Agriculture, Journal Year: 2024, Volume and Issue: 226, P. 109425 - 109425

Published: Sept. 10, 2024

Language: Английский

---

AI-Powered IoT and UAV Systems for Real-Time Detection and Prevention of Illegal Logging

Results in Engineering, Journal Year: 2024, Volume and Issue: unknown, P. 103277 - 103277

Published: Oct. 1, 2024

Language: Английский

---

Improvements in Forest Segmentation Accuracy Using a New Deep Learning Architecture and Data Augmentation Technique

Remote Sensing, Journal Year: 2023, Volume and Issue: 15(9), P. 2412 - 2412

Published: May 5, 2023

Forests are critical to mitigating global climate change and regulating through their role in the carbon water cycles. Accurate monitoring of forest cover is, therefore, essential. Image segmentation networks based on convolutional neural have shown significant advantages remote sensing image analysis with development deep learning. However, learning typically require a large amount manual ground truth labels for training, existing widely used struggle extract details from large-scale high resolution satellite imagery. Improving accuracy remains challenge. To reduce cost labelling, this paper proposed data augmentation method that expands training by modifying spatial distribution images. In addition, improve ability network multi-scale detailed features feature information NIR band images, we high-resolution fusing double input. The experimental results using Sanjiangyuan plateau dataset show our achieves an IoU 90.19%, which outperforms prevalent networks. These demonstrate approaches can forests images more effectively accurately.

Language: Английский

---

Wavelet Transform Cluster Analysis of UAV Images for Sustainable Development of Smart Regions Due to Inspecting Transport Infrastructure

Sustainability, Journal Year: 2025, Volume and Issue: 17(3), P. 927 - 927

Published: Jan. 23, 2025

Sustainable development of the Smart Cities and Regions concept is impossible without a modern transport infrastructure, which must be maintained in proper condition. Inspections are required to assess condition objects infrastructure (OTI). Moreover, efficiency these inspections can enhanced with unmanned aerial vehicles (UAVs), whose application areas continuously expanding. When inspecting OTI (bridges, highways, etc.) problem improving quality image processing, analysis data collected by UAV, for example, particularly relevant. The advanced methods assessing quantity information making decisions reduce uncertainty redundancy such systems often complicated presence noise there. To harmonize characteristics certain procedures conditions, authors propose conducting processing using wavelet transform clustering three main phases: determining number clusters, defining coordinates cluster centres, clustering. We compared existing one transform. research has shown that UAVs used inspecting; moreover, method characterised an improved processing. In addition, assessment enables us degree approximation result ideal one. examined specific challenges associated planning UAV flights during obtain will enhance accuracy recognition. This especially important comprehensive quantitative adaptation tasks “Smart Cities/Regions” based on pragmatic measure informativeness.

Language: Английский

---

NoctuDroneNet: Real-Time Semantic Segmentation of Nighttime UAV Imagery in Complex Environments

Drones, Journal Year: 2025, Volume and Issue: 9(2), P. 97 - 97

Published: Jan. 27, 2025

Nighttime semantic segmentation represents a challenging frontier in computer vision, made particularly difficult by severe low-light conditions, pronounced noise, and complex illumination patterns. These challenges intensify when dealing with Unmanned Aerial Vehicle (UAV) imagery, where varying camera angles altitudes compound the difficulty. In this paper, we introduce NoctuDroneNet (Nocturnal UAV Drone Network, hereinafter referred to as NoctuDroneNet), real-time model tailored specifically for nighttime scenarios. Our approach integrates convolution-based global reasoning training-only alignment modules effectively handle diverse extreme conditions. We construct new dataset, NUI-Night, focusing on low-illumination scenes rigorously evaluate performance under conditions rarely represented standard benchmarks. Beyond assess Varied Dataset (VDD), normal-illumination demonstrating model’s robustness adaptability flight domains despite lack of large-scale Furthermore, evaluations Night-City dataset confirm its scalability applicability urban environments. achieves state-of-the-art surpassing strong baselines both accuracy speed. Qualitative analyses highlight resilience under-/over-exposure small-object detection, underscoring potential real-world applications like emergency landings minimal illumination.

Language: Английский

---

Forestry Segmentation Using Depth Information: A Method for Cost Saving, Preservation, and Accuracy

Forests, Journal Year: 2025, Volume and Issue: 16(3), P. 431 - 431

Published: Feb. 27, 2025

Forests are critical ecosystems, supporting biodiversity, economic resources, and climate regulation. The traditional techniques applied in forestry segmentation based on RGB photos struggle challenging circumstances, such as fluctuating lighting, occlusions, densely overlapping structures, which results imprecise tree detection categorization. Despite their effectiveness, semantic models have trouble recognizing trees apart from background objects cluttered surroundings. In order to overcome these restrictions, this study advances management by integrating depth information into the YOLOv8 model using FinnForest dataset. Results show significant improvements accuracy, particularly for spruce trees, where mAP50 increased 0.778 0.848 mAP50-95 0.472 0.523. These findings demonstrate potential of depth-enhanced limitations RGB-based segmentation, complex forest environments with structures. Depth-enhanced enables precise mapping species, health, spatial arrangements, habitat analysis, wildfire risk assessment, sustainable resource management. By addressing challenges size, distance, lighting variations, approach supports accurate monitoring, improved conservation, automated decision-making forestry. This research highlights transformative integration models, laying a foundation broader applications environmental conservation. Future studies could expand dataset diversity, explore alternative technologies like LiDAR, benchmark against other architectures enhance performance adaptability further.

Language: Английский

---