REAL TIME DETECTION OF FOREST FIRES AND VOLCANIC ERUPTIONS FROMMETEOSAT SECOND GENERATION IMAGES USING A NEURAL NETWORK

摘要

One of the most important parameters in the estimation of the evolution of global change is the gas composition of the atmosphere and its temporal variation. Amongst the vari-ous and complex processes that absorb or produce gases, the biomass burning has very important short and long term ef-fects [ 1 ]. Remote sensing plays a key role in monitoring these effects [2], but you have to make a compromise in temporal, spectral and spatial resolution [3, 4]. As burning savannas represents the main contribution to global biomass burning, monitoring Africa becomes a priority. Because of its near real time imaging capacities and its position over the African Continent, Meteosat Second Generation (MSG) appears to be a very adapted satellite to efficiently do this task[5]. The approach described in this article is based on an un-dergraduate project which test the potentiality of neural net-work for hot spot detection in MSG images. The main authors are the undergraduate student that have achieved this promis-ing project.