Forest Wildfire Detection and Forecasting Utilizing Machine Learning and Image Processing

Wildfires are spontaneous, unwanted, uncontrolled fires in large forest and due to global warming these type of incidents are increasing. These Wildfires causes damage to forest’s natural ecosystems globally, resulting in ecological, environmental and even economic degradation. Hence it is an urgent need to detect or even predict such wildfires before they become large or out of control. This project is an attempt to design a machine learning based wildfires detection and prediction system to detect such wildfire incidents at early stages as well as predict. The ML-based system is planned to use the remote-sensed satellite images of a particular geographical region and use image processing to detect wildfires early, if any. Otherwise, the system would use the other parameters related to the region in the image like temperature, wind speed, pressure, vegetation, etc. to predict the occurrence of wildfires in near future. The supervised machine learning algorithms like kNN, Logistic Regression, and Random Forest will be applied to train and test the large sets of images and data. A comparative study of the algorithms would be done and depending on the results of the study, the best of them would be selected for use by the system. The system will have a front-end web interface where any user (like forest monitoring officials, environmental protection agencies, etc..) can upload data for any particular geographical region (like temperature, wind speed, pressure etc.) along with it’s remote-sensed satellite image for either detecting or predicting the possibility of wildfires in that region of concern.