Automatic extraction of building roofs by bayesian analysis of laser scanneing data: potentials of appication in sensor pose estimation

• Grant number: 05/04645-0
• Support Opportunities: Regular Research Grants
• Start date: April 01, 2006
• End date: March 31, 2008
• Field of knowledge: Physical Sciences and Mathematics - Geosciences - Geodesy
• Principal Investigator: Aluir Porfírio Dal Poz
• Grantee: Aluir Porfírio Dal Poz
• Host Institution: Faculdade de Ciências e Tecnologia (FCT). Universidade Estadual Paulista (UNESP). Campus de Presidente Prudente. Presidente Prudente , SP, Brazil
• City of the host institution: Presidente Prudente

Abstract

Road network automatic extraction from multi-sensor data (laser scanning and high-resolution multi-spectral aerial and satellite images) is the recent research focus of my group. Laser scanning data may be efficiently used in the roof contour extraction, whose results are fundamental for the extraction of two important road network contextual elements: shadows and perspective obstructions of buildings onto streets. This contextual information is fundamental for facilitating the complex image analysis operations in high-resolution images, as such one that allows a detailed description of the urban road network. Three general problems may be identified: 1- extraction of contextual information of road network using mainly laser scanning data; 2- integration of multi-sensor data; and 3- automatic extraction of the urban road network using high-resolution images. The solution of the two first problems would be used in assisting the solution of the third problem. In this project it is proposed the investigation of two subproblems related to the first two general problem mentioned above: extraction of roof contours using laser scanning data; and 2- digital image orientation in relation to the laser scanning data reference system. Concerning the first subproblem, it is proposed a methodology for automatic extraction of building roof contours in two main steps: region segmentation of a DEM (Digital Elevation Model) obtained from laser scanning data; and recognizing of the segmented regions corresponding to the roof contours using MRF (Markov Random Fields) models. Concerning the second subproblem, it is proposed the evaluation of a photogrammetric model for aerial image orientation in relation to the laser scanning reference system using the roof contour's straight lines as ground control... (AU)