Segmentação de telhados de edifícios em dados de varredura a LASER usando técnicas de processamento de imagem e o método RANSAC

• Ano de defesa: 2015
• Autor(a) principal: Yano, Michelle Sayuri [UNESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Cartografia; Edificios; Telhados; Processamento de imagens; Lasers; Amostragem (Estatistica); Cartography
• Link de acesso: http://hdl.handle.net/11449/149242; http://www.athena.biblioteca.unesp.br/exlibris/bd/cathedra/17-01-2017/000876648.pdf

This research proposes a method for segmentation of buildings roof planes present in LASER scanning data, based on image processing techniques and the RANSAC (RANdom SAmple Consensus) method. To avoid the exhaustive search of planar roof face segments throughout the LASER point cloud, to decrease significantly the computational effort and avoid the merge of different building segments accidentally coplanar, firstly, a point cloud preprocessing is performed in order to separate and acquire the points belonging to each building. This preprocessing detects the points belonging to roof regions and then, delimits the region of each building from the contours acquired by detection, vectorization and polygonization of the building edges. Once acquired the building points, the RANSAC method is used to detect sets of coplanar points belonging to roofs faces. The principal differential of this research is the development of a method for automatic determination of the thresholds used in RANSAC method during the stage of consistency analysis between the data and the hypothetical model obtained...