Forest fire and smoke detection using deep learning-based learning without forgetting

Fire Ecology, Journal Year: 2023, Volume and Issue: 19(1)

Published: Feb. 17, 2023

Abstract Background Forests are an essential natural resource to humankind, providing a myriad of direct and indirect benefits. Natural disasters like forest fires have major impact on global warming the continued existence life Earth. Automatic identification is thus important field research in order minimize disasters. Early fire detection can also help decision-makers plan mitigation methods extinguishing tactics. This looks at fire/smoke from images using AI-based computer vision techniques. Convolutional Neural Networks (CNN) type Artificial Intelligence (AI) approach that been shown outperform state-of-the-art image classification other tasks, but their training time be prohibitive. Further, pretrained CNN may underperform when there no sufficient dataset available. To address this issue, transfer learning exercised pre-trained models. However, models lose abilities original datasets applied. solve problem, we use without forgetting (LwF), which trains network with new task keeps network’s preexisting intact. Results In study, implement such as VGG16, InceptionV3, Xception, allow us work smaller lessen computational complexity degrading accuracy. Of all models, Xception excelled 98.72% We tested performance proposed LwF. Without LwF, among gave accuracy 79.23% (BowFire dataset). While 91.41% for BowFire 96.89% dataset. find fine-tuning LwF performed comparatively well Conclusion Based experimental findings, it found current methods. show successfully categorize novel unseen datasets.

Language: Английский

---

Assessing the Effectiveness of YOLO Architectures for Smoke and Wildfire Detection

IEEE Access, Journal Year: 2023, Volume and Issue: 11, P. 96554 - 96583

Published: Jan. 1, 2023

This paper presents a comprehensive evaluation of various YOLO architectures for smoke and wildfire detection, including YOLOv5, YOLOv6, YOLOv7, YOLOv8, YOLO-NAS. The study aims to assess their effectiveness in early detection wildfires using the Foggia dataset, comprising 8,974 images specifically designed this purpose. Performance employs metrics such as Recall, Precision, F1-score, mean Average Precision provide holistic assessment models' performance. follows rigorous methodology involving fixed epochs, continuous performance tracking, unbiased testing. Results show that YOLOv8 exhibit balanced across all both validation YOLOv6 performs slightly lower recall during but achieves good balance on YOLO-NAS variants excel recall, making them suitable critical applications. However, precision is models. Visual results demonstrate top-performing models accurately identify most instances test set. they struggle with distant scenes poor lighting conditions, occasionally detecting false positives. In favorable perform well identifying relevant instances. We conclude no single model excels aspects detection. choice depends specific application requirements, considering accuracy, inference time. research contributes field computer vision providing foundation improving systems mitigating impact wildfires. Researchers can build upon these findings propose modifications enhance systems.

Language: Английский

---

Leveraging the power of internet of things and artificial intelligence in forest fire prevention, detection, and restoration: A comprehensive survey

Internet of Things, Journal Year: 2024, Volume and Issue: 26, P. 101171 - 101171

Published: March 26, 2024

Forest fires are a persistent global problem, causing devastating consequences such as loss of human lives, harm to the environment, and substantial economic losses. To mitigate these impacts, accurate prediction early detection forest is critical. In response this challenge living in digital era Artificial Intelligence (AI) smart economies, there has been growing interest utilising AI mechanisms for fire management. This study provides an in-depth examination use algorithms fight against fires. particular, our paper starts with overview followed by comprehensive review various systems approaches. includes thorough analysis works that have evaluated factors influence occurrence severity, well those focus on systems. The also explores adapting restoring after concludes evaluation potential impact management suggestions future research directions, taking full advantage novel technologies, 5G communications, Software Defined Networking (SDN), twins, federated learning blockchain. Finally, draws lessons insights limitations management, highlighting need further development field maximise its benefits.

Language: Английский

---

Inference of Wildfire Causes From Their Physical, Biological, Social and Management Attributes

Earth s Future, Journal Year: 2025, Volume and Issue: 13(1)

Published: Jan. 1, 2025

Abstract Effective wildfire prevention includes actions to deliberately target different causes. However, the cause of an increasing number wildfires is unknown, hindering targeted efforts. We developed a machine learning model ignition across western United States on basis physical, biological, social, and management attributes associated with wildfires. Trained from 1992 2020 12 known causes, overall accuracy our exceeded 70% when applied out‐of‐sample test data. Our more accurately separated ignited by natural versus human causes (93% accuracy), discriminated among 11 classes human‐ignited 55% accuracy. attributed greatest percentage 150,247 for which source was unknown equipment vehicle use (21%), lightning (20%), arson incendiarism (18%).

Language: Английский

---

Assessing forest fire likelihood and identification of fire risk zones using maximum entropy-based model in Khyber Pakhtunkhwa, Pakistan

Environmental Monitoring and Assessment, Journal Year: 2025, Volume and Issue: 197(3)

Published: Feb. 13, 2025

Language: Английский

---

Assessment of the performance of GIS-based analytical hierarchical process (AHP) approach for flood modelling in Uttar Dinajpur district of West Bengal, India

Geomatics Natural Hazards and Risk, Journal Year: 2022, Volume and Issue: 13(1), P. 2183 - 2226

Published: Aug. 19, 2022

Floods have received global significance in contemporary times due to their destructive behavior, which may wreak tremendous ruin on infrastructure and civilization. The present research employed an integration of the Geographic information system (GIS) Analytical Hierarchy Process (AHP) method for identifying flood susceptibility zonation (FSZ), vulnerability (FVZ), risk (FRZ) humid subtropical Uttar Dinajpur district India. study combined a large number thematic layers (N = 12 FSZ N 9 FVZ) achieve reliable accuracy included multicollinearity analysis these variables overcome issues related highly correlated variables. According findings, 27.04, 15.62, 4.59% area were classified as medium, high, very high FRZ, respectively. ROC-AUC, MAE, MSE, RMSE model exhibited good prediction 0.73, 0.15, 0.16, 0.21, performance AHP has been evaluated using sensitivity analyses. It also recommends that persistent improvement this subject, such studies modifying criteria thresholds, changing relative criteria, desired matrix, will permit GIS MCDA be progressively adapted real hazard-management issues.

Language: Английский

---

Deep Learning Approaches for Wildland Fires Remote Sensing: Classification, Detection, and Segmentation

Remote Sensing, Journal Year: 2023, Volume and Issue: 15(7), P. 1821 - 1821

Published: March 29, 2023

The world has seen an increase in the number of wildland fires recent years due to various factors. Experts warn that will continue coming years, mainly because climate change. Numerous safety mechanisms such as remote fire detection systems based on deep learning models and vision transformers have been developed recently, showing promising solutions for these tasks. To best our knowledge, there are a limited published studies literature, which address implementation classification, detection, segmentation As such, this paper, we present up-to-date comprehensive review analysis methods their performances. First, previous works related including reviewed. Then, most popular public datasets used tasks presented. Finally, discusses challenges existing works. Our shows how approaches outperform traditional machine can significantly improve performance detecting, segmenting, classifying wildfires. In addition, main research gaps future directions researchers develop more accurate fields.

Language: Английский

---

FirePred: A hybrid multi-temporal convolutional neural network model for wildfire spread prediction

Ecological Informatics, Journal Year: 2023, Volume and Issue: 78, P. 102282 - 102282

Published: Sept. 1, 2023

Language: Английский

---

Forest fire management, funding dynamics, and research in the burning frontier: A comprehensive review

Trees Forests and People, Journal Year: 2024, Volume and Issue: 16, P. 100526 - 100526

Published: Feb. 29, 2024

We indexed 8,970 scientific publications on forest fires in order to bridge the gap between research and policy discussions fires. Journal articles conference papers dominated literature, with an emphasis environmental science, agricultural biological sciences, earth planetary engineering, computer science. Research field of fire has historically focused terms such as "Forest Fire", "Wildfire", "Deforestation", but recent trends have highlighted "MODIS," "Artificial Intelligence," "Algorithm," "Satellite Data," "Prediction.". The number steadily risen, particularly after 2000, funding predominantly from National Science Foundation, Natural U.S. Forest Service, Aeronautics Space Administration. Notable contributions observed United States, China, Canada, Spain, Australia, India. International Wildland had maximum share published among journals, followed by Ecology Management, Forests, Total Environment, Remote Sensing. A variety aspects been covered, data-driven studies, new discoveries, methodological advances, theoretical applications, governance implications. In spite our long interrelation fires, we are lacking a comprehensive mechanism combat them effectively. multidisciplinary approach collection analysis information could provide insightful tool for evidence-based policies practices aimed address emerging challenges due at global scale.

Language: Английский

---

Comparing machine learning algorithms to predict vegetation fire detections in Pakistan

Fire Ecology, Journal Year: 2024, Volume and Issue: 20(1)

Published: June 24, 2024

Abstract Vegetation fires have major impacts on the ecosystem and present a significant threat to human life. consists of forest fires, cropland other vegetation in this study. Currently, there is limited amount research long-term prediction Pakistan. The exact effect every factor frequency remains unclear when using standard analysis. This utilized high proficiency machine learning algorithms combine data from several sources, including MODIS Global Fire Atlas dataset, topographic, climatic conditions, different types acquired between 2001 2022. We tested many ultimately chose four models for formal processing. Their selection was based their performance metrics, such as accuracy, computational efficiency, preliminary test results. model’s logistic regression, random forest, support vector machine, an eXtreme Gradient Boosting were used identify select nine key factors and, case vegetation, seven that cause fire findings indicated achieved accuracies ranging 78.7 87.5% 70.4 84.0% 66.6 83.1% vegetation. Additionally, area under curve (AUC) values ranged 83.6 93.4% 72.6 90.6% 74.2 90.7% model had highest accuracy rate also AUC value proving be most optimal model. provided predictive insights into specific conditions regional susceptibilities occurrences, adding beyond initial detection data. maps generated analyze Pakistan’s risk showed geographical distribution areas with high, moderate, low risks, highlighting assessments rather than historical detections.

Language: Английский

---