Laser Scanning Data Segmentation in Urban Areas by a Bayesian Framework

• Autor(a) principal: Galvanin, Edinéia Aparecida dos Santos
• Data de Publicação: 2007
• Outros Autores: Dal Poz, Aluir Porfírio, Pires de Souza, Aparecida Doniseti
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Boletim de Ciências Geodésicas
• Texto Completo: https://revistas.ufpr.br/bcg/article/view/8246
• topic: Markov Random Field, Modelo Digital de Elevação, Segmentação por região, QuadTree; Markov Random Field, Digital Elevation Model, Region Segmentation

In this paper is presented a region-based methodology for Digital Elevation Modelsegmentation obtained from laser scanning data. The methodology is based on twosequential techniques, i.e., a recursive splitting technique using the quad treestructure followed by a region merging technique using the Markov Random Fieldmodel. The recursive splitting technique starts splitting the Digital Elevation Modelinto homogeneous regions. However, due to slight height differences in the DigitalElevation Model, region fragmentation can be relatively high. In order to minimizethe fragmentation, a region merging technique based on the Markov Random Fieldmodel is applied to the previously segmented data. The resulting regions are firstlystructured by using the so-called Region Adjacency Graph. Each node of theRegion Adjacency Graph represents a region of the Digital Elevation Modelsegmented and two nodes have connectivity between them if corresponding regionsshare a common boundary. Next it is assumed that the random variable related toeach node, follows the Markov Random Field model. This hypothesis allows thederivation of the posteriori probability distribution function whose solution isobtained by the Maximum a Posteriori estimation. Regions presenting highprobability of similarity are merged. Experiments carried out with laser scanningdata showed that the methodology allows to separate the objects in the DigitalElevation Model with a low amount of fragmentation.