CLASSIFICATION ORBITAL IMAGE FOR MAXIMUM LIKELIHOOD METHOD IN QUIXERAMOBIM, CEARA, BRAZIL/CLASSIFICACAO DE IMAGEM ORBITAL PELO METODO MAXIMA VEROSSIMILHANCA EM QUIXERAMOBIM, CEARA, BRASIL/CLASIFICACION DE IMAGEN ORBITAL POR EL METODO DE MAXIMA VEROSIMILITUD EN QUIXERAMOBIM, CEARA, BRASIL

• Published in: Revista geográfica acadêmica 2016-01, Vol.10 (1), p.81
• Main Authors: Brasileiro, Felipe Gomes, de Oliveira, Carlos Magno Moreira, de Avila Rodrigues, Rafael, Coll Delgado, Rafael
• Format: Article
• Language: Portuguese
• ISSN: 1678-7226; 1678-7226

This study evaluated the effectiveness of supervised maximum likelihood algorithm (MAXVER) in the area covered by the municipality of Quixeramobim, Ceara, Brazil, using the Landsat 8 satellite images and geo tools for use mapping and land cover. To measure as ground truth we used the GPS (GARMIN ETrex LEGEND Cx), sampling the different classes of use and land cover. The classes sampled in the field were: urban area, exposed soil, grassland, savanna and water. The results showed that the overall rating of the image had a reasonable value, indicating an overall accuracy of 57.50%, which indicates a greater need for sample data collection for hazardous area. The Kappa when observed by class, showed low values for grassland soil and exposed areas (14% and 30%), very low kappa indices obtained showing that the sample was unsuitable for classification of the spectral response of these features. Classes urban area and water (50% and 54%) had median values of Kappa for his features with respect to sampling for the classification of their spectral response. However, the savanna class was the one that presented representative value according to the Kappa index (77%). For future studies, it is recommended to use sensors/orbital platforms that have a higher spatial resolution, temporal and spectral. Keywords: Geotechnology, image classification, orbiting platforms. Este trabalho avaliou a eficiencia do algoritmo Maxima Verossimilhanca (MAXVER), do municipio de Quixeramobim, Ceara, Brasil, utilizando imagens do satelite Landsat 8 sensor OLI para o mapeamento do uso e cobertura do solo. Para a afericao como verdade de campo utilizou-se o GPS (GARMIN Etrex LEGEND Cx), amostrando as diferentes classes de uso e cobertura do solo. As classes amostradas em campo foram: area urbana, solo exposto, pastagem, caatinga e agua. Os resultados encontrados mostraram que a classificacao geral da imagem teve um valor razoavel, indicando acuracia global de 57,50%, o que indica uma maior necessidade de coleta de dados de amostras para a area classificada. O indice Kappa, quando observado por classe, apresentou valores baixos para areas de solo exposto e pastagem (14% e 30%), obtiveram baixissimos indices Kappa mostrando que a amostragem foi inadequada para classificacao da resposta espectral destas feicoes. As classes area urbana e agua (50% e 54%) apresentaram valores medianos de Kappa para suas feicoes em relacao a amostragem para a classificacao de sua resposta espectral. No entanto,