Resumen

PALACIO PRIETO, José Luis  y  LUNA GONZALEZ, Laura. Clasificación espectral automática vs. Clasificación visual: Un ejemplo al sur de la Ciudad de México. Invest. Geog [online]. 1994, n.29, pp.25-40. ISSN 2448-7279.

Using a Maximum Likelihood algorithm a Landsat TM image was classified by both supervised and non-supervised approaches. In the first case, 12 classes were obtained based on 30 samples; the non-supervised procedure yielded 30 classes. Once grouped, both classifications considered 6 classes. Additionally, color composites were prepared and visually interpreted. The three products were compared in a GIS environment using a regularly distributed network of points refering the field truth. The results show that the lowest error correspond to the supervised classification (82.32% exactitude), followed by the visual interpretation (78.72%) and the non-supervised procedure (73.18%). These figures were obtained after grouping the classes according to their similarities.