Implementación de redes neuronales convolucionadas para la clasificación de imágenes de sensores remotos en la Amazonía

Abstract

The monitoring of the land surface through remote sensing has been driven by the need to know the territory and, thus, make optimal decisions when managing natural resources and manage the sustainable use of human activities. Given the recent advances in the acquisition and processing of data from remote sensors on different platforms with different characteristics in terms of geometric, radiometric, spectral and temporal resolutions, it has been possible to collect information from large areas of the earth's surface, thus generating large volumes of information that require analysis in real time to maximize the use of the knowledge obtained. This is how data science since the 80s has developed the methods of deep learning and computational vision as artificial neural networks and specifically convoluted neural networks CNN for the capture, processing and analysis of images showing good results in the identification of objects. This project offers a CNN model for the classification of land cover in the Amazon using images from the PLANET platform.