Automatic extraction of building roof contours by laser scanning data and markov random field

• Autor(a) principal: Galvanin, Edinéia Aparecida dos Santos
• Data de Publicação: 2008
• Outros Autores: Dal Poz, Aluir Porfírio, Souza, Aparecida Doniseti Pires de
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Boletim de Ciências Geodésicas
• Texto Completo: https://revistas.ufpr.br/bcg/article/view/11817
• topic: Automatic Extraction; Building Roof Contours; Digital Elevation Model; Laser Scanning Data; Markov Random Field; Extração Automática; Contornos de Telhados de Edifícios; Modelo Digital de Elevação; Dados de Varredura a Laser; Campos Randômicos de Markov

This paper proposes a methodology for automatic extraction of building roofcontours from a Digital Elevation Model (DEM), which is generated through theregularization of an available laser point cloud. The methodology is based on twosteps. First, in order to detect high objects (buildings, trees etc.), the DEM issegmented through a recursive splitting technique and a Bayesian mergingtechnique. The recursive splitting technique uses the quadtree structure forsubdividing the DEM into homogeneous regions. In order to minimize thefragmentation, which is commonly observed in the results of the recursive splittingsegmentation, a region merging technique based on the Bayesian framework isapplied to the previously segmented data. The high object polygons are extracted byusing vectorization and polygonization techniques. Second, the building roofcontours are identified among all high objects extracted previously. Taking intoaccount some roof properties and some feature measurements (e. g., area,rectangularity, and angles between principal axes of the roofs), an energy functionwas developed based on the Markov Random Field (MRF) model. The solution ofthis function is a polygon set corresponding to building roof contours and is foundby using a minimization technique, like the Simulated Annealing (SA) algorithm. Experiments carried out with laser scanning DEM´s showed that the methodologyworks properly, as it delivered roof contours with approximately 90% shapeaccuracy and no false positive was verified.