Inteligencia computacional en teledeteccióncontrol de contenido de humedad en combustible en superficie terrestre mediante imagen satélite para prevención de incendios

Abstract

This doctoral thesis examines the problems related to the estimation of fuel moisture content to prevent possible fires and how to behave once they commence. The thesis describes the estimation problem using Remote Sensing techniques. We have achieved a breakthrough in terms of research into new unconventional computational models to estimate the amount of water in different environments of vegetation, one of the biggest problems of the Remote Sensing, because it requires a prior study of the area, collecting selective treatment of these samples and to understand their structure, composition and moisture content. Through these simulations we have analyzed the capacity of discrimination by analyzing the ROC curves (Receiver Operating Characteristic), and later use different Soft Computing techniques, namely, neural and evolutionary models, to validate the experiments. The methodology presented in this paper to estimate moisture content in vegetation is based on the use of radiative transfer models by linking models at Prospect leaf (leaf optical spectra properties) (Jacquemoud et al. 1990) and canopy model SAIL (Scattering by Arbitrarily Inclined Leaves).