Klasifikácia Land Cover v využitím umelých neurónových sietí

Abstract

The aim of this thesis is the classification of satellite images using artificial neural networks in order to recognize the types of surfaces for the creation of land cover maps. The introduction describes the methods of supervised and unsupervised classification with emphasis on artificial neural networks. The practical part describes the implementation of supervised and unsupervised classification and testing of artificial neural networks with the evaluation of the accuracy of each output. Landsat 8 images were classified. At the end of the work individual methods of classification are compared. Classification by supervised and unsupervised classification took place in the PCI Geomatica program and classification by artificial neural networks in the SAGA GIS program.