Fuzzy logic and neural network approaches for land cover mapping of the Brazilian territory

Abstract

Land cover mapping is important for regional and urban plannings. Most of the digital remote sensing based
mappings use conventional classification techniques which are limited in their ability to define a wide variety
of the existing types of land cover. The objective of this paper is to analyze the potential of fuzzy logic and
artificial neural network approaches to produce land cover maps of the Brazilian territory for regional scales
and urban environments. Two case studies were developed to test the methodological approach. Reliability
analysis of resulting thematic maps was conducted through statistical coefficients and uncertainty
measurements. Results showed that the use of fuzzy logic and artificial neural network approaches are
suitable to fulfill the land cover mapping needs of the Brazilian territory.