A Review of Wildfire Detection: Methodologies and Comparison Analysis

Due to the havoc cause to ecosystems, human lives, and property, wildfires are a major environmental concern on a worldwide scale. Using Unmanned Aerial Vehicles (UAVs) and satellite-based technologies are the two main approaches that are the subject of this paper's review of the research on wildfire detection strategies. UAV-based detection systems provide early wildfire identification and real-time monitoring by utilizing state-of-the-art technology like computer vision and deep learning. Numerous studies demonstrate the effectiveness of Unmanned Aerial Vehicles (UAVs) fitted with a range of models, such as YOLOv3, YOLOv5, and distinct Convolutional Neural Networks (CNNs), attaining elevated detection accuracy levels between 71.72% and 99.20%. In contrast, satellite-based methods leverage advanced deep learning models like U-Net, DeepLabV3+, and CNNs to precisely detect and map wildfires using datasets from platforms like Landsat-8, Sentinel-2, and Himawari-8. The techniques have remarkable efficacy, as evidenced by F1-scores of up to 98.78% and a 94% F1-score for smoke detection using geostationary satellites. This thorough analysis highlights the difficulties and developments in wildfire detection technologies, highlighting the critical roles that machine learning, remote sensing, and creative sensor systems play in reducing the harmful consequences of wildfires and facilitating prompt response actions.