Automatic Extraction of Facade Details of Heritage Building Using Terrestrial Laser Scanning Data

摘要

The Terrestrial Laser Scanning (TLS) has become an interesting technology in various fields. The architecture and heritage fields benefit more and more of this technology. In the heritage field, TLS data and derived products are used to complete the digital archive of cultural sites all over the world. Among these cultural sites, we deal with the historical Moroccan sites, especially the old Medina. The rehabilitation of the old Medina requires the reconstruction of facade planes. For that purpose, Moroccan authorities need a precise inventory of all facades. The planes extracted from 3D point cloud provide the desired results. However, the manual process is too long and sometimes difficult. Automatic methods of plane recognition including the point cloud segmentation are generally based on geometric approaches which fail to identify some facade details of such heritage buildings. In this context, we propose a new automatic approach of point cloud segmentation. This approach relies on all the components of a colored point--both geometric and radiometric--combining the RGB values, laser intensity and geometric data. Our approach also includes a new method to filter the segmentation result through Delaunay triangulation. The last step of our processing is the facade and detail contour detection that is based on alpha-shape algorithm to find interior and exterior boundaries. Experiments are performed on facades presenting an example of old Medina architecture located in Casablanca’s Medina. Results show the importance of integrating all point cloud components for the detail facades extraction and planes establishment.