Drone Assisted Deep Learning based Wildfire Detection System

Wildfires are one of the major threats to forest sustainability. Therefore, there is an urgent need for an early prevention system for wildfires. The flight controller comprises a Node Micro-Controller Unit board and is controlled by an ESP8266-based Wireless setup. To address the problem of wildfire monitoring in forests, this paper proposes a wildfire monitoring system based on drones and deep learning techniques. In the present study, after pre-processing the images collected from a drone camera, deep learning models are used to classify them as containing fire or non-fire. Among the approaches discussed, the customized Inception model trained over the combined data-set from six different sources outperforms the previously presented works, achieving an overall validation accuracy of 96.04% and performing well on fire-like images and from bright day to dim light images also.