Implementación del método máquinas de soporte vectorial en bases de datos espaciales para análisis de clasificación supervisada en imágenes de sensores remotos

Abstract

The present project is oriented to the implementation of a supervised classification method on images from remote sensors stored in a spatial database that allows contributing to the diagnosis of image classification, according to parameters of normality and abnormality where it is also possible to store these results within the same database manager system. Given that the supervised classification algorithm Vector Support Machines (MSV) is widely accepted as one of the best classification techniques because it allows to have a very good accuracy in the diagnosis of the different coverages present in the ground, since it seeks not only to find a dissociation between these, but to achieve a separation between the elements to be classified, will be implemented as a classification technique for the pilot project to be carried out. The application will be designed for the end user, which allows not only to obtain support and sustenance when making decisions, but also to facilitate the updating of the database, the inclusion or elimination of information from it, as well as the possibility to choose the main characteristics that must be taken into account during the classification process. This utility is of great value, since when working with images of similar characteristics, the possibility of establishing dissociation ranges or weights to the different coverages directly affects the expected result. Finally, a case study related to the deforestation of the Colombian Amazon will be presented, where the usefulness of the application will be demonstrated through a supervised classification which will be compared with the classification module of some software that implements it at present.