Automatic Building Reconstruction from Cadastral Maps and Aerial Images

The approach described in this paper is intended for 3D building reconstruction on large and dense urban areas. It provides a real-time solution for automatic inferring of polyhedral rooftops of buildings from cadastral maps and aerial images. From the initial planimetric outline and hypotheses of planes inferred from it, a set of rooftop models depending on two parameters (slope and gutter altitude) is extracted through maximal cliques enumeration methodology. Among this family, the best representation is chosen by centered correlation on Digital Elevation Model computed by correlation from aerial images. Finally, slope and altitude are determined by L1 minimization. Some results on real data show the validity of the algorithm that favors robustness against generality and is designed for integration in a semi-automatic platform for 3D urban databases constitution.