PyroScan: Wildfire Behavior Prediction System using Unmanned Aerial Vehicles (UAV)

During a wildfire, it is of the utmost importance to be updated about all information of the wildfire. Wind speed, wind direction and dry grass often works as fuel for the fire allowing it to spread in multiple directions. These different factors are often issues for any firefighting organization that is trying to help fight the fire. An uncontrolled wildfire is often a threat to wildlife, property, and worse, human and animal lives. In our paper, we propose a computer vision powered fire tracking and prediction application utilizing Unmanned Aerial Vehicles (UAV) to inform fire fighters regarding the probability of the direction of a fire. Our work is finally aimed to assist firefighters to know which front would need to be attended first based on wind speed, wind direction, dry grass, etc. The UAV will use the powers of a mounted camera and computer vision to capture and calculate the current trajectory and rate of spread from an overhead live feed of the fire. Currently, we use the powers of classic Computer Vision techniques to identify a fire, and then using Kalman Filtering, track the direction of it. Based on our results, we believe that this work can generate meaningful information which will help firefighters protect property damage and even save lives.