The majority of existing infrastructure buildings are not recorded digitally. Considering that most of the projects in this sector are refurbishments and expansion tasks the current state of the assets needs to be captured and processed into a suitable digital model. New improved capturing techniques, like laser scanning, provide the surveying data for this task in the form of point clouds. The manual analysis of the vast amount of point cloud data which contain billions of points is a tedious and error-prone endeavour. This thesis deals with the automated processing of those data sets with the intention of producing semantically rich geometric tunnel models. The implemented algorithms are capable of producing satisfactory results in labelling data points into different categories like lining, catenary, objects, walkways and trackbed. Simple model fitting is applied to the segmented point portions. Furthermore, expansions as well as emergency doors are recognised. The rudimentary tunnel data structure defined in this work as well as the chosen segmentation methods can be extended for future efforts in creating a building information model for tunnel infrastructures.     «

The majority of existing infrastructure buildings are not recorded digitally. Considering that most of the projects in this sector are refurbishments and expansion tasks the current state of the assets needs to be captured and processed into a suitable digital model. New improved capturing techniques, like laser scanning, provide the surveying data for this task in the form of point clouds. The manual analysis of the vast amount of point cloud data which contain billions of points is a tedious a...     »