Fusion of Terrestrial and Airborne Laser Data for 3D Modeling Applications

Abstract

This thesis deals with the 3D modeling phase of the as-built large BIM projects. Among several means of BIM data capturing, such as photogrammetric or range tools, laser scanners have been one of the most efficient and practical tool for a long time. They can generate point clouds with high resolution for 3D models that meet nowadays’ market demands. The current 3D modeling projects of as-built BIMs are mainly focused on using one type of laser scanner data, such as Airborne or Terrestrial. According to the literature, no significant (few) efforts were made towards the fusion of heterogeneous laser scanner data despite its importance. The importance of the fusion of heterogeneous data arises from the fact that no single type of laser data can provide all the information about BIM, especially for large BIM projects that are existing on a large area, such as university buildings, or Heritage places. Terrestrial laser scanners are able to map facades of buildings and other terrestrial objects. However, they lack the ability to map roofs or higher parts in the BIM project. Airborne laser scanner on the other hand, can map roofs of the buildings efficiently and can map only small part of the facades. Short range laser scanners can map the interiors of the BIM projects, while long range scanners are used for mapping wide exterior areas in BIM projects. In this thesis the long range laser scanner data obtained in the Stop-and-Go mapping mode, the short range laser scanner data, obtained in a fully static mapping mode, and the airborne laser data are all fused together to bring a complete effective solution for a large BIM project. Working towards the 3D modeling of BIM projects, the thesis framework starts with the registration of the data, where a new fast automatic registration algorithm were developed. The iii next step is to recognize the different objects in the BIM project (classification), and obtain 3D models for the buildings. The last step is the development of an occlusion removal algorithm to efficiently retain parts of the buildings occluded by surrounding objects such as trees, vehicles, or street poles.