Data-driven forest fire analysis

Abstract

There are typically about 100,000 wildfires in the United States every year. Over 9 million acres of land have been destroyed due to treacherous wildfires. It is difficult to predict and detect Forest Fire in a sparsely populated forest area and it is more difficult if the prediction is done using ground based methods like Camera or Video Based approach. Satellites can be an important source of data prior to and during the Fire due to its reliability and efficiency. This paper discusses various real-time forest fire detection and prediction approaches with a goal of informing the local fire authorities. The algorithms used in fire detection consider the geometry as well as the characteristics of the images generated.