Desarrollo y propuesta metodológica para el empleo de los Campos Aleatorios de Markov aplicados a técnicas de clasificación de coberturas en imágenes de la superficie terrestre

Abstract

Although the Markov Random Fields (MRF) theory is widely accepted by the scientific community as one of the best modeling techniques for random variables processes, and has been widely used in the areas of physics, mathematics, image processing and computer vision, its capabilities have not been considered for a key process in remote sensing as it is the classification of land covers. This thesis is oriented to the implementation of a supervised classification method of multispectral optical images of earth observation, within an application that allows the classification of image patterns.An application is presented that is designed for the end user, which allows not only technically and theoretically support the user when making decisions, but also facilitates the possibility of choosing and changing the parameters that must be taken into account during the classification process according to the selected method. This application is of great value, since, when working with multispectral images, the possibility of viewing them and being able to select the samples of the different land covers and perform a rapid validation in the same application is beneficial for anyone interested in digital image processing and remote sensing. It was found that the classification of images using MRF allows to have a good accuracy in the diagnosis of the different coverages present in the soil, since it achieves a separation between the elements to be classified, which will be implemented as a classification technique for the investigation. Finally, a validation and comparison of the method studied with respect to classical methods (KNN and Malahanobis) of classification where the utility of the application will be demonstrated.