Comportamento de sistemas de informações geográficas por meio de classificação supervisionada em diferentes bacias hidrográficas

• Ano de defesa: 2015
• Autor(a) principal: Rodrigues, Mikael Timóteo [UNESP]
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Geoprocessamento; Sensoriamento remoto; Bacias hidrograficas; Sistemas de informação geografica; Solos - Uso; Remote sensing
• Link de acesso: http://hdl.handle.net/11449/135950; http://www.athena.biblioteca.unesp.br/exlibris/bd/cathedra/10-02-2016/000858808.pdf

The aim of this study is to determine the three GIS software behavior (version IDRISI Selva, ArcGIS 10.1 and TerraView 4.2.2) through supervised classification by spectral pattern on Landsat 5, associated with the comparison of land use in the basins Lavapés the river and Capybara, Botucatu / SP, using techniques of remote sensing and GIS. The areas of supervised training were defined from nine classes for basin Lavapés and seven for basin Capybara, fundamental for the study and analysis of the use and occupation of land, proposed by the Manual Use of Technical IBGE Earth. To identify the best classification, were crossed the maximum likelihood maps (MAXVER) derived from Geographic Information Systems with the ground reality, where it is characterized as the actual land use, pointing accuracy (accuracy) of each classification, crossing Pixel arrays or sets of pixels. There were significant difference of data obtained from the supervised classification by maximum likelihood generated in the software IDRISI Selva, ArcGIS 10.1 and TerraView 4.2.2. The difference in results between the two evaluated basins was significant, where the basin of Capybara presented in all GIS software best results safely by having a smaller number of land use classes and a smaller urban area, thus causing less confusion for the algorithm. Another obvious factor was the difference of products derived from the supervised classification by maximum likelihood generated in software and then post classified with the majority filters (MAJORITY FILTER), where ever after reclassification accuracy was high, presented smaller error omission and commission in the ...