Use of Airborne Laser Scanning Data and Image-based Three-dimensional (3-D) Edges for Automated Planar Roof Reconstruction


DEMİR N.

LASERS IN ENGINEERING, vol.32, no.3-4, pp.173-205, 2015 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 32 Issue: 3-4
  • Publication Date: 2015
  • Journal Name: LASERS IN ENGINEERING
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.173-205
  • Keywords: Airborne laser scanning, LiDAR, roof reconstruction, three-dimensional (3-D) modelling, point cloud, data fusion, object extraction, BUILDING EXTRACTION, LIDAR DATA, AERIAL, LINES
  • Akdeniz University Affiliated: Yes

Abstract

In this work an automated approach for 3-D roof reconstruction using airborne images and LiDAR data is presented. The roofs are assumed to consist of planar parts. Reconstruction is thus conducted using plane detection with random sample consensus (RANSAC), followed by a geometric refinement and merging of the neighbouring segments to reduce over segmentation. Planes belonging to the walls and ground are then detected and excluded, and the roof outlines are vectorized using the alpha-shapes algorithm. The resulting polygons are refined using straight three-dimensional (3-D) edges reconstructed through automatic edge extraction and matching, along with virtual 3-D edges from the intersection of adjacent planes. Additionally, 3-D corner points constructed through the intersection of the 3-D lines are used. The 3-D edges and corner points are integrated with the point clouds belonging to the planes, and the roof outlines are again detected using the alpha-shapes algorithm. Finally, a regularisation of the roof outlines is performed, although the topological correctness of the roof planes is not enforced. The quantitative evaluation uses several metrics for both the correctness and completeness of the roof polygons and for their geometric accuracy. The average mean and median value for the correctness of the roof polygons are 84 and 96%, respectively, and for completeness are 76 and 88%. In addition, different cases and problems are analysed and visually presented.