AbstractAirborne Laser Scanning is an advanced surveying technology (also called Light Detection and Ranging - LIDAR) which mounts a laser scanner on an aircraft. The aircraft scans the Earth's surface and captures data by emitting and receiving light pulses transmitted onto the terrain objects. The captured data are in three dimensions (3D); however no extra information is provided to describe them. Additional algorithms are needed to extract meaningful and useful information from the data.
The popularity of LIDAR has attracted attention as researchers try to develop algorithms for 3D building reconstruction in Geographical Information Systems. The limited information provided by the data makes building boundary and roof structure extractions become essential tasks when analysing the data.
This research examines the limitations of different algorithms for extracting building outlines and remodelling roof structures from the LIDAR data solely, and suggests an alternative approach for reconstructing buildings using raw LIDAR data. Most of the current methods use additional data sources (e.g. cadastral data, aerial photos, or satellite images) and pre-defined building
models to reconstruct 3D buildings.
The extraction method proposed starts by re-sampling the captured data in a lower resolution index layer and the aim is to search for vertical wall segments which separate the high and low areas. The wall segments found are connected and modified to form closed building outlines and corners. The roof remodelling system suggested starts by creating a triangulation using the extracted data points which are inside the building boundaries. Three clustering methods are used to separate the triangles into groups which share the same properties (e.g. orientation and geographical location). Each group of triangles represents a plane on the roof. Plane to plane relationships are found, and the building corners and roof ridges are calculated by using the three planes intersection. Finally the building is reconstructed from the terrain model using a set of well-developed toolkits to extend the TIN model with preserved topological connectivity.
Real LIDAR data are used to evaluate the capability and the validity of the developed algorithms. The data were captured in Bournemouth by the Ordnance Survey UK. In conclusion, several suggestions are made to improve the algorithms for future development.
|Date of Award