ACURÁCIA TEMÁTICA DO CLASSIFICADOR POR MÁXIMA VEROSSIMILHANÇA EM IMAGEM DE ALTA RESOLUÇÃO ESPACIAL DO SATÉLITE GEOEYE-1.

This study evaluates the thematic accuracy of the maximum likelihood classifier in a high spatial resolution imaging satellite GeoEye-1. The study area refers to the sub-basin of the stream Horizonte, Alegre, ES and the classes were analyzed agriculture area, urban area, forest fragmentation, hydrography, pasture and bare soil. The quality of the thematic map was assessed using the kappa statistic, overall accuracy, producer's accuracies and user, with reference to the mapping of classes in the scale of 1:2,000 mapping. The results show that automatic classification given by the classifier gives good results for kappa (58.34%) and overall accuracy (71.41%). Among the classes evaluated, the fragment forest and pasture were those with the best accuracies and precisions. The recognition of other classes as agriculture area, urban area, hydrography and exposed soil, depending on the complexity of the landscape and its small dimensions in the study area, depends on the use of image interpretation techniques for further details. [ABSTRACT FROM AUTHOR]