Deep Learning based Forest Fire Detection and Alert System

Abstract

A Wildfire or Forest Fire that originates within a woodland or any forest area gives rise to air pollution. Since the burning releases huge quantities of carbon di-oxide, carbon monoxide and fine particulate matter into the atmosphere, it causes enormous damage to the vegetation and wildlife in the nearby area. In this paper, a Deep Learning based Convolutional Neural Network (CNN) model is proposed to detect forest fire. In the proposed work, the following techniques are used: Image Collection, Pre-processing and Image Classification. Initially, the images in the dataset are pre-processed, and fed into the CNN for feature extraction and detection. Further, the hardware setup is implemented using Raspberry Pi, in which the fire detection alerts are sent as an e-mail, buzzer and LCD display to the concerned authorities with detection accuracy of 93% and 92% on training and testing datasets, respectively.