Comparison of supervised classification methods of Maximum Likelihood, Minimum Distance, Parallelepiped and Neural Network in images of Unmanned Air Vehicle (UAV) in Viçosa - MG

Abstract

The aim of this work was testing the classification techniques in digital aerial images of spatial high resolution obtained by Unmanned Air Vehicle (UAV). The images recover an area of the Federal University of Viçosa, campus Viçosa in the municipality of Minas Gerais, Brazil. From the orthophoto generated, the classification test was made, by using four classifiers: Maximum Likelihood, Minimum Distance, Parallelepiped and Neural Network. The classification that best delimited the different features present in the image was the classification by Artificial Neural Networks. In order to prove statistically the classification efficiency, the validation was carried out through Kappa index and visual analysis.