Network of Drones for Early Warning of Forest Fire and Dynamic Fire Quenching Plan Generation

Abstract

Abstract Wildfires, one of the most frequent natural disasters, significantly harm society, the economy, and the environment. Transfer learning algorithm and modern machine learning tools can help in early forest fire prediction, detection, and dynamic fire quenching. Also, a dynamic fire quenching plan is prescribed which takes into account the live status of forest fire. A network of drones which has high definition image processing and decision making capabilities are used for detection of forest fire in the very early stage there by preventing the forest fire. In case fire breaks out, the proposed system generates a fire quenching plan and alerts the fire and rescue departments so that they do plan early and stop the forest fire as soon as possible. ResNet, VGGNet, Mobile Net, Alexi Net, GoogLeNet models enhanced with transfer learning are used to predict the forest fire events in India. It is found that the proposed technique GoogLeNet-TL, which uses transfer learning with GoogLeNet has resulted in 96% accuracy and 97.56% F1 score than other models considered for forest fire prediction in Indian subcontinent.