@Article{AlmeidaFoNascBati:1998:TMImLa,
               author = "Almeida Filho, Raimundo and Nascimento, Paulo S{\'e}rgio Rezende 
                         and Batista, Getulio Teixeira",
          affiliation = "{Instituto Nacional de Pesquisas Espaciais (INPE)} and {Instituto 
                         Nacional de Pesquisas Espaciais (INPE)} and {Instituto Nacional de 
                         Pesquisas Espaciais (INPE)}",
                title = "Avalia{\c{c}}{\~a}o de t{\'e}cnicas de segmenta{\c{c}}{\~a}o 
                         e classifica{\c{c}}{\~a}o autom{\'a}tica de imagens Landsat-TM 
                         no mapeamento do uso do solo na Amaz{\^o}nia / Evaluation of 
                         segmentation and automatic classification techniques of landsat - 
                         TM imagery for land use mapping in Amazonia",
              journal = "Acta Amazonica",
                 year = "1998",
               volume = "28",
               number = "1",
                pages = "41--53",
             keywords = "mapeamento tem{\'a}tico automatizado, segmenta{\c{c}}{\~a}o de 
                         imagens, classifica{\c{c}}{\~a}o n{\~a}o supervisionada por 
                         regi{\~o}es, mudan{\c{c}}as no uso da terra, sensoriamento 
                         remoto, automated thematic mapping, image segmetation, per-field 
                         non-supervised classification, land use change, remote sensing.",
             abstract = "0 mapeamento do uso da terra e fundamental para o entendimento dos 
                         processos de mudan{\c{c}}as globais, especialmente em 
                         regi{\~o}es como a Amaz{\^o}nia que est{\~a}o sofrendo grande 
                         press{\~a}o de desenvolvimento. Tradicionalmente estes 
                         mapeamentos tem sido feitos utilizando t{\'e}cnicas de 
                         interpreta{\c{c}}{\~a}o visual de imagens de sat{\'e}lites, 
                         que, embora de resultados satisfat{\'o}rios, demandam muito tempo 
                         e alto custo. Neste trabalho e proposta uma t{\'e}cnica de 
                         segmenta{\c{c}}{\~a}o da imagens com base em um algoritmo 
                         decrescimento de regi{\~o}es, seguida de uma 
                         classifica{\c{c}}{\~a}o n{\~a}o-supervisionada por 
                         regi{\~o}es. Desta forma, a classifica{\c{c}}{\~a}o 
                         tem{\'a}tica se refere a um conjunto de elementos (pixels da 
                         imagem), beneficiando-se portanto da informa{\c{c}}{\~a}o 
                         contextual e minimizando as limita{\c{c}}{\~o}es das 
                         t{\'e}cnicas de processamento digital baseadas em analise pontual 
                         (pixel-a-pixel). Esta t{\'e}cnica foi avaliada numa {\'a}rea 
                         t{\'{\i}}pica da Amaz{\^o}nia, situada ao norte de Manaus, AM, 
                         utilizando imagens do sensor {"}Thematic Mapper{"} - TM do 
                         satelite Landsat, tanto na sua forma original quanto decomposta em 
                         elementos puros como vegeta{\c{c}}{\~a}o verde, 
                         vegeta{\c{c}}{\~a}o seca (madeira), sombra e solo, aqui 
                         denominada imagem nusturas. Os resultados foram validados por um 
                         mapa de referencia gerado a partir de t{\'e}cnicas consagradas de 
                         interpreta{\c{c}}{\~a}o visual, com verifica{\c{c}}{\~a}o de 
                         campo, e indicaram que a classifica{\c{c}}{\~a}o autom{\'a}tica 
                         e vi{\'a}vel para o mapeamento de uso da terra na Amaz{\^o}nia. 
                         Testes estat{\'{\i}}sticos indicaram que houve concord{\^a}ncia 
                         significativa entre as classifica{\c{c}}{\~o}es autom{\'a}ticas 
                         digitais e o mapa de refer{\^e}ncia (em tomo de 95 de 
                         confian{\c{c}}a). ABSTRACT: Land use mapping is essential for the 
                         understanding of global change processes, especially in regions 
                         such as the Amazon that are experiencing great pressure for 
                         development. Traditionally, these mappings have been done using 
                         visual interpretation techniques applied to satellite imagery. 
                         These techniques provide satisfactory results but are 
                         time-consuming and very costly. In the present paper, a technique 
                         is proposed that uses image segmentation based on an algorithm for 
                         expansion of homogeneous regions on the image; application of the 
                         algorithm is followed by a non-supervised region-by-region 
                         classification. Thus, the thematic classification is based on a 
                         set of image elements (pixels), benefiting from contextual 
                         information, thereby avoiding the limitations of digital 
                         processing techniques that are based on single pixels (per-pixel 
                         classification). This approach was evaluated in a typical test 
                         site in the Amazon region located to the north of Manaus, 
                         Amazonas, using both original Landsat Thematic Mapper images and 
                         their decomposition into fractions of endmembers such as green 
                         vegetation, woody material, shade and soil, called mixed images in 
                         this paper. The results were validated against a reference map 
                         obtained from proven techniques for visual interpretation of 
                         satellite imagery and by field checking. The results indicate that 
                         mapping land use in Amazonia using automatic classification is 
                         feasible. Statistical tests indicated that there was significant 
                         agreement between the automated digital classifications and the 
                         reference map (at the 95% confidence level).",
                 issn = "0044-5967",
                label = "8458",
             language = "pt",
           targetfile = "artigo.pdf",
        urlaccessdate = "03 jun. 2024"
}