Method for the automated generation of a forest non forest map with LANDSAT 8 imagery by using artificial neural networks and the identification of pure class pixels

Abstract

In this work, a methodology for the automated classification of Landsat 8 images from the integration of Artificial Neural Networks and the identification of pixels of pure classes is presented. The exercise carried out in this research by using the SEPAL platform, allowed to obtain a mosaic L8 of the study area, fully preprocessed and calibrated, and it was generated automatically in a short period of time. This result represents a significant advance in terms of preprocessing capacity that currently exists for the management of satellite data compared to the state of the area a decade ago. This relevant advance has been possible due to the use of artificial neural networks and the cross-correlation coefficient of the pixels of the Landsat 8 satellite platform images. Their use and differentiation of areas in remote sensing of wooded, agricultural and water areas are discussed.