Fire Detection Using Support Vector Machines (SVM)

Earth is an intrinsically flammable planet owing to its cover of carbon-rich vegetation, seasonally dry climates, wide spread lightning and volcano ignitions. Therefore, something is always burning on this earth. Wildfires can be natural or man-made (accidental or deliberate). These fires release greenhouse gases and huge volumes of smoke thereby polluting the air and the environment which in turn cause severe degradation to the ecosystems. Based on observations from the Moderate Resolution Imaging Spectroradiometer (MODIS) on NASAs Terra and Aqua satellites, several algorithms have been designed and many approaches have been proposed to detect wildfire. While the question of accuracy over the existing methods of fire detection remains, this paper proposes a more robust and accurate approach using SVM (Support Vector Machines) for detecting wildfire. And also will compare SVM with logistic regression which is the most common used algorithm for solving industry problems.