Jost, Berit Henrike: Strategies for the Empirical Determination of the Stochastic Properties of Terrestrial Laser Scans. - Bonn, 2023. - Dissertation, Rheinische Friedrich-Wilhelms-Universität Bonn.
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5-71297
@phdthesis{handle:20.500.11811/10913,
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5-71297,
author = {{Berit Henrike Jost}},
title = {Strategies for the Empirical Determination of the Stochastic Properties of Terrestrial Laser Scans},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2023,
month = jun,

volume = 919,
note = {Terrestrial laser scanners (TLS) are suitable for the surface approximation of objects and their geometric changes due to the temporally and spatially high-frequent data acquisition. However, precise geodetic engineering tasks require detailed knowledge about the performance of the sensors and especially about their uncertainty to use them for precise measurements, e.g., in deformation analysis or surface approximations.
Due to the complex transition behavior between error sources and effects on the point cloud, the correct description of the point cloud's stochastic model represented by the variance-covariance matrix is not yet solved. In particular, the interaction of the laser beam with the environment and the measurement object, taking into account different measurement arrangements (distances and angles of incidence), is so diverse that it is impossible to model all errors. However, if these errors are neglected in the stochastic model, this can lead to biased surface approximations, incorrect statistical tests, or misinterpreting errors as deformations. For this reason, strategies for the empirical determination of the stochastic properties of terrestrial laser scans are developed in this dissertation. In particular, the determination of the range precision for different surfaces and measurement configurations, as well as correlations between individual measurement points, are in focus. Specifically, the following aspects are addressed:
The object surface and scanning configuration mainly influence the range precision, which the reflected intensity of the laser beam can fully describe. This work contributes to efficiently determining the range precision by presenting a test field simplified for users and further developing the existing methodology. This contributes to a more realistic description of the main diagonal of the variance-covariance matrix representing the stochastic model.
Especially the interaction of the laser beam with the object is individual as it depends on the surface. The laser spot is integrated over a certain area, and neighboring laser spots overlap due to the dense acquisition of data points. This results in a smoothing effect and leads to the fact that the resolution capability of the scanner does not match the resolution set in the scanner. This thesis develops a new method for determining the resolution capability, which enables a more economic measurement planning. Furthermore, correlations are derived from overlapping laser spots, which are integrated into the stochastic model.
These rather short-scale correlations can be determined empirically via another method developed in this thesis. For this purpose, the stochastic signal of the point cloud must first be separated from the deterministic part. This is done with the help of a reference geometry generated with a sensor of higher accuracy. Subsequently, this work presents two methods for quantifying the short-scale correlations in the point cloud.
The previous methods can be implemented well for point clouds of smaller objects (approximately up to 2 m x 2 m). However, this is not straightforward to realize for larger objects as the stochastic properties change within the point cloud. Furthermore, a reference geometry is not easy to establish due to a lack of suitable sensors and deformations of the reference objects. For this reason, this thesis presents a method for creating a reference geometry of a larger object that allows for the analysis of long-scale correlations.
These different aspects provide a better understanding of the uncertainties in terrestrial laser scanning and, thus, form the basis for setting up a more realistic stochastic model of the point cloud to make statistically more reliable statements in a deformation analysis and unbiased surface approximations. Furthermore, the presented strategies do not require special laboratory conditions but can be performed by qualified users if an appropriate object, such as a roughly planar wall, is available.},

url = {https://hdl.handle.net/20.500.11811/10913}
}

The following license files are associated with this item:

InCopyright