Comparación de las técnicas de clasificación de imágenes satelitales J48, PART mediante el uso de componentes principales

Abstract

A fundamental tool for mapping is spatial remote sensing, a technique that allows the acquisition of multispectral images from remote sensors. These images provide a greater number of observations of the coverages, than any terrestrial survey, working at the same scale. Over the years, public availability and image costs were more accessible, coupled with advances in both spatial and spectral resolution. For this reason their interpretation and analysis are in increasing development. ' The present work tries to use aerial images as sources of data, through which a thematic and logical classification of the same is obtained. The data in a satellite image can be classified in coverage of soil, water and even temperature. For this reason, the present thesis will begin observing the definitions of remote sensors and aerial images in order to learn their potentialities and advantages. Then it will be presented how an aerial image can be decomposed into vectors which hold information that will serve as input to the neural network. As each vector holds important data, each one is specialized in one type of data.