Reconstrução de contornos e cumeeiras de telhados de edifícios a partir de imagens aéreas de alta-resolução e poliedros representativos de edifícios extraídos de dados LASER

• Ano de defesa: 2013
• Autor(a) principal: Marcato, Vanessa Jordão [UNESP]
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual Paulista (Unesp)
• Palavras-chave em Português: Cartografia; Fotogrametria; Imagens digitais; Cartography
• Link de acesso: http://hdl.handle.net/11449/108622

This paper proposes improvements in a previous methodology for the geometric refinement of building roof contours extracted from LASER data using high-resolution aerial images and Markov Random Field (MRF) models. The original methodology takes for granted that the 3D description of each building roof reconstructed from the laser scanning data is topologically correct and that it is only necessary to improve its accuracy. Since roof ridges are accurately extracted from laser scanning data, our main objective is to use high-resolution aerial images to improve the accuracy of roof outlines. For this, this methodology uses lines extracted form the image and the projection of the roof contours extracted from the LASER data to establish a MRF description based on relations of length, proximity and orientation between the two sets of straight lines. One of the proposed improvements is to include in the energy function associated with the MRF a restriction called corner injuction. This restriction considers that the intersection of two adjacent lines, corresponding to the contour sides, should be close to a corner in 90º. The corners are extracted from an image through an appropriate image processing algorithm. Other improvement in the energy function is based on the fact that the lines representing roof contours are aproximately parallel or orthogonal. This restriction was called rectangularity injuction. Other modification in the original methodology refers to the energy function optimization algorithm. The original methodology used the brute force optimization algorithm associated with some heuristics. Although this method allows to obtain the optimal solution, if there is, the search space becomes computationally intractable when dozens of lines are in the search space. It is proposed to use a genetic algorithm in order to solve this problem...