2019-09-052019-09-052018-02-16GOMES, Paulo Isaac Moura de. Análise orientada a objetos do sensor óptico ASTER aplicadas à classificação de cangas lateríticas na região de Carajás. Orientador: Arnaldo de Queiroz da Silva. Coorientador: Wilson da Rocha Nascimento Júnior. 2018. 92 f. Trabalho de Curso (Bacharelado em Geologia) - Faculdade de Geologia, Instituto de Geociências, Universidade Federal do Pará, Belém, 2018. Disponível em: http://bdm.ufpa.br/jspui/handle/prefix/2020. Acesso em:.http://bdm.ufpa.br/jspui/handle/prefix/2020The evolution of the multispectral images obtained in the near infrared and short wave infrared fields of orbital sensors greatly increased the discrimination capacity of the terrestrial targets. One of the privileged areas of application with this type of image is the remote sensing of the rock-soil-vegetation interactions, which, despite already having sources of information with high spatial resolution, still had data restriction that incorporated better spectral and radiometric resolutions. The combination of these two characteristics allowed to improve the detection of elements that compose a landscape allowing applications in the mapping of vegetation cover, geological and soil use, from the aid of automatic classifiers. However, the application of automatic classification techniques to images from the multispectral medium-resolution spatial resolution systems has encountered some difficulties because the spatial and radiometric resolution of these images do not have the ability to separate small targets, therefore, their intra- and inter- classes may be limited. Thus, classifiers that use pixel-to-pixel-based methods are restricted to classify these types of images because they work only with spectral information, which is not always sufficient to discriminate features with a high variety of responses. The distinction of classes in this type of images can occur through the inclusion of other attributes / information, such as shape, size and context in the classification. The incorporation of these attributes into the classifiers defines object-oriented analysis and is an option to overcome the limitation of pixel-to-pixel classifiers, considering that topological (neighborhood, context) and geometric information (shape and size) are used in the classification process. The present work aims to explore object - oriented analysis in the classification of lateritic cangas in the region of the Carajás Mineral Province, from multispectral images in the fields of visible and near infrared (VNIR) and short - wave infrared (SWIR), using an ASTER image. The study area includes the municipalities of Canaã dos Carajás, Parauapebas and Curionópolis. For that, an experiment was performed in which three classes were individualized: vegetation, lateritic canga, exposed soil. The classification involved multiresolution segmentation, definition of a hierarchical network for classification of the objects, validation of the results reaching a global accuracy index of 91.33% and a total Kappa index of 87%. Thus, it was possible to conclude that the object-oriented classification method for mapping the classes showed expressive results, based on spectral attribute information and customized attributes, with the possibility of finding specific thresholds of the work targets, and for the creation of an unpublished band math for the automatic classification of lateritic cangas of the region of Carajás.Acesso AbertoSensoriamento remotoClassificadores automáticosAnálise orientada a objetosCangas lateríticasProvíncia Mineral de Carajás - PACNPQ::CIENCIAS EXATAS E DA TERRA::GEOCIENCIAS::GEOLOGIATrabalho de Curso - Graduação - Monografia