DEVELOPING FIRE DETECTION ALGORITHMS BY GEOSTATIONARY ORBITING PLATFORMS AND MACHINE LEARNING TECHNIQUES

摘要

Fires in general and forest fires specific are a majorrnconcern in terms of economical and biological loses.rnRemote sensing technologies have been focusing onrndeveloping several algorithms, adapted to a large kindrnof sensors, platforms and regions in order to obtainrnhotspots as faster as possible. The aim of this study is tornestablish an automatic methodology to develop hotspotsrndetection algorithms with Spinning Enhanced Visiblernand Infrared Imager (SEVIRI) sensor on board MeteosatrnSecond Generation platform (MSG) based on machinernlearning techniques that can be exportable to othersrngeostationary platforms and sensors and to any area ofrnthe Earth. The sensitivity (SE), specificity (SP) andrnaccuracy (AC) parameters have been analyzed in orderrnto develop the final machine learning algorithm takingrninto account the preferences and final use of thernpredicted data.