@InProceedings{FerreiraZorZanShiSou:2015:MaEsAr,
               author = "Ferreira, Matheus Pinheiro and Zortea, Maciel and Zanotta, Daniel 
                         Capella and Shimabukuro, Yosio Edemir and Souza Filho, Carlos 
                         Roberto de",
          affiliation = "{Instituto Nacional de Pesquisas Espaciais (INPE)} and {} and {} 
                         and {Instituto Nacional de Pesquisas Espaciais (INPE)}",
                title = "Mapeamento de esp{\'e}cies arb{\'o}reas em floresta tropical 
                         utilizando imagens hiperespectrais",
            booktitle = "Anais...",
                 year = "2015",
               editor = "Gherardi, Douglas Francisco Marcolino and Arag{\~a}o, Luiz 
                         Eduardo Oliveira e Cruz de",
                pages = "1224--1230",
         organization = "Simp{\'o}sio Brasileiro de Sensoriamento Remoto, 17. (SBSR)",
            publisher = "Instituto Nacional de Pesquisas Espaciais (INPE)",
              address = "S{\~a}o Jos{\'e} dos Campos",
             abstract = "Tree species mapping in tropical forests using remote sensing 
                         imagery is challenging with the benefit to provide valuable 
                         insights for ecologists and land managers. Hyperspectral data 
                         proved to be feasible for this task, but it is still unclear how 
                         different classification methods perform. In this work, we 
                         evaluated Linear Discriminant Analysis (LDA), Radial Basis 
                         Function Support Vector Machines (RBF-SVM) and Random Forest (RF) 
                         for tree species discrimination and mapping in a tropical forest 
                         using airborne hyperspectral data. The effects of dimensionality 
                         reduction on classification performance were also assessed by 
                         selecting sets of 10, 20 and 30 bands. At the pixel level, LDA 
                         performed better than other methods (Average Accuracy (AA) =84.7%) 
                         using all (260) spectral bands for classification. However, 
                         RBF-SVM produced the best map of species using 30 selected bands 
                         (AA =90.4%) in an object based approach (OBIA). OBIA increased the 
                         AA of species mapping for all tested methods and reduced spatial 
                         noise.",
  conference-location = "Jo{\~a}o Pessoa",
      conference-year = "25-29 abr. 2015",
                 isbn = "978-85-17-0076-8",
                label = "226",
             language = "pt",
         organisation = "Instituto Nacional de Pesquisas Espaciais (INPE)",
                  ibi = "8JMKD3MGP6W34M/3JM47SC",
                  url = "http://urlib.net/ibi/8JMKD3MGP6W34M/3JM47SC",
           targetfile = "p0226.pdf",
                 type = "Sensoriamento remoto hiperespectral",
        urlaccessdate = "27 abr. 2024"
}