Large-scale deformation monitoring using mobile laser scanning and tailored point cloud processing

This dissertation introduces a novel method for deformation analysis of retaining structures using mobile laser scanning (MLS). The idea is to use a commercial MLS system on the roof of a vehicle to capture infrastructure while passing by, thus without impeding traffic flow. New, tailored algorithms enable the analysis of the MLS data in an automated, efficient, and repeatable manner. The goal is to apply the approach on a large scale to detect those objects that show signs of damage among the vast number of structures. The tailored point cloud processing algorithms work for different structural types and surface properties. In short, the algorithms aim to describe complex structural deformation patterns by rigid-body motion of many small parts. These small parts may represent perceptually meaningful objects or point subsets created by spatial clustering. Both methods yield groups of thousands of points, allowing precise and robust estimates of tilt and displacements. This thesis presents novel methods to eliminate systematic discrepancies between two MLS point clouds. The benefits are significant, as laser scans from two different epochs can be registered over stable reference surfaces. Moreover, it allows determining uncertainty of MLS point clouds. Both aspects, accurate registration and knowledge about uncertainty, are crucial components of deformation monitoring. Finally, the thesis presents a guideline for implementing the proposed methodology in practice. The guideline addresses the needs of both infrastructure operators and service providers. The aim is to avoid monopolies while ensuring the quality of the deformation analysis independent of the service providers.