@InProceedings{ParreiraDutPanRuwLu:2015:MéCoCl,
               author = "Parreira, Michelle de Oliveira and Dutra, Luciano Vieira and 
                         Pantale{\~a}o, Eliana and Ruwer, Sherfis Gibran and Lu, 
                         Dengsheng",
          affiliation = "{Instituto Nacional de Pesquisas Espaciais (INPE)} and {Instituto 
                         Nacional de Pesquisas Espaciais (INPE)} and {} and {Instituto 
                         Nacional de Pesquisas Espaciais (INPE)}",
                title = "Sistema Classificador Parreira: um m{\'e}todo de 
                         combina{\c{c}}{\~a}o de classifica{\c{c}}{\~o}es por pares de 
                         classes",
            booktitle = "Anais...",
                 year = "2015",
               editor = "Gherardi, Douglas Francisco Marcolino and Arag{\~a}o, Luiz 
                         Eduardo Oliveira e Cruz de",
                pages = "2929--2936",
         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 = "Traditional procedures for classifying remote sensing images use 
                         the concept of competitive classification. It chooses the 
                         classification that achieves the best results among the tested 
                         metrics. The problem within this procedure is the loss of 
                         information for some class when just one classifier is chosen, 
                         since each classifier generates different sampling error. This 
                         paper presents the first part of the develop-ment of a new 
                         Classification System, called Parreira. It combines the results of 
                         classifiers to analyze the discriminability of class pairs. From 
                         an image, a set of classes and their training ROIs, the system 
                         gener-ates all possible combinations of classes in pairs. By JM 
                         distance, it selects the three attributes that allow greater 
                         discriminalidade between each class pair and performs 
                         classification of them. For this paper, just the Maximum 
                         Likelihood and Support Vector Machine classifiers were used, in a 
                         single hierarchical level. The resulting classification is made by 
                         taking classes that were more often identified by the classifier 
                         within subsets of class pairs. The class pairs classifications 
                         showed better separability when compared to a classification of 
                         all classes at the same time. This result shall be studied to 
                         prove its validity or if it is due to the inability to correctly 
                         classify pixels belonging to classes not involved in pair 
                         classification.",
  conference-location = "Jo{\~a}o Pessoa",
      conference-year = "25-29 abr. 2015",
                 isbn = "978-85-17-0076-8",
                label = "585",
             language = "pt",
         organisation = "Instituto Nacional de Pesquisas Espaciais (INPE)",
                  ibi = "8JMKD3MGP6W34M/3JM4AGS",
                  url = "http://urlib.net/ibi/8JMKD3MGP6W34M/3JM4AGS",
           targetfile = "p0585.pdf",
                 type = "Classifica{\c{c}}{\~a}o e minera{\c{c}}{\~a}o de dados",
        urlaccessdate = "27 abr. 2024"
}