|M.Sc Student||Mor Matan|
|Subject||A Hierarchal Approach to Integrating Airborne LiDAR and|
2D GIS for Creating 3D GIS
|Department||Department of Civil and Environmental Engineering||Supervisor||Professor Emeritus Yerach Doytsher|
|Full Thesis text - in Hebrew|
The requirement for accurate 3D buildings models is growing fast among different GIS users in various fields. Communication analysis, 3D cadaster management, 3D visualization and more are various applications based on 3D GIS infrastructure.
This research presents a new and fully automatic method of a 3D reconstruction of buildings based on fusion of LiDAR dataset with 2D polygonal footprint. 2D polygons' footprints are available, from cadastral data and from vector layers. These layers are usually created manually by skilled operators, and as a result this data is very reliable with high accuracy. 2D polygons provide explicit and definite boundary information which makes the crucial fusion process easier.
A hierarchic methodology was developed in order to produce 3D buildings under topological and geometric constraints. The suggested methodology contains four fundamental steps: (1) Global adjustment: when fusing different kinds of data, there is a need to overcome mismatching. An images process algorithm based on cross correlation, and mechanism of adjustment between tow graphs of buildings' centroid was developed, thus enabling to find the registration parameters between the datasets. (2) Classification and filtering the LiDAR data: points are divided into separate groups of ground and non-ground points. Roof's points are classified as object points by a region growing process. (3) Local adjustment: in order to reconstruct the 2D building footprint, a local adjustment of roof's points and the 2D building polygon is essential. Local adjustment is based on the assumption that an optimal adjustment can be achieved when maximal laser points fall inside the building polygon. (4) 3D reconstruction: After extracting edge's points using the "gift wrapping algorithm", a process of linear least square adjustment of piecewise functions is established in order to reconstruct roof's edge.
This methodology was tested over synthetic data and then over real datasets in the cities of Stuttgart and Vaihingen in Germany. While the LiDAR data was taken from the ISPRS website, the GIS buildings' layer was taken from other source. Applying the method on real datasets, even though the complexity of roof tops, the achieved results were still relatively very good. The reconstruction process preserves topology and symmetries of man-made objects. The process results of Stuttgart's buildings reconstruction achieved only a limited level of success due to very complex roofs. Even though, roofs tops as hip, gable and flat were reconstructed in high fidelity in all cases.