Authors: Chaskar Siddhi Dattatray, Chinchawade Tanvi Santosh, Bhore Tanvi Santosh, Prof. P. S. Dumbre, Prof. M. S. Nalawade, Dr. V. M. Dhede
Abstract: Forest fires destroy trees, wildlife habitats and natural resources, release large amounts of smoke and threaten nearby communities. Their impact depends strongly on how quickly they are detected, yet traditional methods based on watchtowers and ground patrols require continuous human effort and are slow to cover remote terrain. Satellites observe large areas but revisit a location only a few times a day and cannot see small fires under cloud or canopy. Unmanned aerial vehicles, commonly called drones, fill this gap by carrying cameras and sensors over the forest on demand. This survey reviews forest fire detection methods and focuses on drone-based systems. It discusses the sensing payload, including visible and thermal cameras and gas and temperature sensors, the detection algorithms, from colour-rule methods to convolutional neural networks, and the communication links that deliver alerts with GPS coordinates to forest authorities. Representative studies and datasets are compared. A layered architecture for a smart drone-based fire detection and alert system is then presented, covering sensing, onboard processing, communication, ground-station verification and alert dissemination. Practical challenges such as limited flight time, false alarms, communication range and regulation are discussed along with future directions such as drone swarms and integration with satellite data.
DOI: https://doi.org/10.5281/zenodo.23862991
