Analysis of Deep Learning Methods for Early Wildfire Detection Systems: Review
Ahmed Saleem Mahdi, Sawsen Abdulhadi Mahmood
Abstract
A wildfire is an uncontrollable fire that arises outside of a specific focus, damages property, and poses a threat to human life and health. Thousands of fires are started every year for a variety of reasons, including dry seasons, thunderstorms, and volcanic activity. However, in recent years, the human factor has emerged as the main reason for irreparable forest fires. In order to address the problem of wildfire detection, deep learning model-based wildfire detection and recognition have been attracted by researchers. Deep learning models using Convolutional Neural Network (CNN) are essentially required several dataset samples to train the network with high accuracy. Meanwhile, the common wildfire datasets are small, repetitious and collected as images or videos samples from the Internet and experiments and there’s no standard dataset to adopt in the training mode and performance evaluation task. This paper presents a literature review of fire detection methods based on Computer Vision systems. It’s focused on recent machine learning models used in addition to the datasets required to construct future research projects for the wildfire detection system. Further, a comparison study is performed to highlight the strength and weaknesses key points of current methods including detection accuracy.