University of Natural Resources and Life Sciences, Vienna (BOKU) - Research portal

Logo BOKU Resarch Portal

Gewählte Doctoral Thesis:

Doctoral Thesis - Institut für Alpine Naturgefahren (IAN), BOKU-Universität für Bodenkultur, pp 75. UB BOKU obvsg FullText

Data Source: ZID Abstracts
A digital elevation model (DEM) is the basis for many scientific and engineering applications. The relatively new technology of laser scanning offers new possibilities in terms of speed, accuracy and safety when recording terrain surfaces. Before a DEM can be created, the non-ground points must be separated from the ground points by filtering. Most existing filtering algorithms assume recording data obtained from the airborne laser scanning (ALS) method. Since the terrestrial laser scanning (TLS) measurement is from a static center, the spatial data distribution of these data is very different from ALS data. In this thesis, new approaches are presented, which incorporate the special properties of a TLS recording into the filtering process. Such an approach is based on the assumption that there can be no other ground point over a direct line of sight between the laser scanner and ground point. The final goal was to develop a procedure for filtering of terrestrial laser scan data as complete as possible. The filtering should not only be carried out by the Z value (height), as is usual with many existing filter algorithms, but also laterally using length values. Through this combination of spatial information from different directions, non-ground points should be properly detected and removed even in areas with very few original data. For this purpose, the idea was to use the horizontal angle values and the vertical angle value instead of the usual X, Y point values, resulting in a very uniform grid without large holes. Recording length is then used as funtion value instead of the Z point value. By combining these two sets of filters, the statistical quality of the DEM generated could be significantly increased in any case. The presented methods could also help to expand existing filtering methods.

Betreuer: Prokop Alexander
1. Berater: Fuchs Helmut

© BOKU Wien Imprint