Fuzzy-fusion approach for land cover classification

The use of computational intelligent techniques for feature extraction and classification from earth observation satellite images, like Landsat multispectral images, can contribute to improve remote sensing analysis. Image fusion techniques are applied to fuse the spectral images into a higher-level image of the land cover distribution. In this paper we propose a fuzzy-fusion inference approach for satellite image classification based on a fuzzy process, which uses both a hybrid method to train the classifier and reinforcement aggregation operators in the inference scheme. The approach was tested with land cover maps for the district of Mandimba of the Niassa province, Mozambique and was validated against an expert classification and then with Decision trees and Artificial Neural Networks.