%0 Conference Proceedings
%4 dpi.inpe.br/marte2/2013/05.28.22.34.28
%2 dpi.inpe.br/marte2/2013/05.28.22.34.29
%@isbn 978-85-17-00066-9 (Internet)
%@isbn 978-85-17-00065-2 (DVD)
%F 115
%T Análise de classificadores para mapeamento de uso e cobertura do solo
%D 2013
%A Augusto-Silva, Pétala Bianchi,
%A Valério, Larissa Patrício,
%A Santos, Thiago Batista dos,
%A Alcântara, Enner Herenio de,
%A Stech, José Luiz,
%@affiliation
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%@affiliation
%@affiliation Instituto Nacional de Pesquisas Espaciais (INPE)
%@electronicmailaddress petala@dsr.inpe.br
%E Epiphanio, José Carlos Neves,
%E Galvão, Lênio Soares,
%B Simpósio Brasileiro de Sensoriamento Remoto, 16 (SBSR)
%C Foz do Iguaçu
%8 13-18 abr. 2013
%I Instituto Nacional de Pesquisas Espaciais (INPE)
%J São José dos Campos
%P 2424-2430
%S Anais
%1 Instituto Nacional de Pesquisas Espaciais (INPE)
%X Land Use and Land Cover (LULC) maps have been developed in order to guide decision-making upon spatial data. This allows the construction of indicators for assessing the support capacity of the environment. In the present work, a map of land cover and use was developed with a TM-Landsat-5 image beyond a multispectral classification scheme. Methods of classification available in the software SPRING were tested to evaluate their performance in organize the study area based on five thematic classes defined by FAO. A reference map was developed by an interpreter based on visual classification aided with high-resolution images from Google Earth. Sample points were selected for comparison between the reference map and the ones automatically classified, and the performance of the classifying methods was evaluated based on the percentage of rights. The Isoseg has the best percentage (83,79%) of rights, but since it is a non-supervised classificator, it separates the scene into much more themes than the others, so we have to remap these themes into the original classes of FAO. Thats why Bhattacharya was considered the best method with 77,25% of rights. A class named antrophic was the one with the worst performance probably because some objects can get mixed up with the spectral response of soil prepared for cultivation. The methods in general can perform well, but in fact they should be used with the aid of interpreter knowledge and knowing the precision its required for the work.
%9 Classificação e Mineração de Dados
%@language pt
%3 p0115.pdf