University of Twente Student Theses


CAD model reconstruction from the LiDAR scan of the catenary arch

Przadka, Bartosz (2022) CAD model reconstruction from the LiDAR scan of the catenary arch.

[img] PDF
Abstract:The railway industry sees a chance to improve its operations both in terms of security and efficiency by digitalizing the railway tracks and nearby structures. Strukton Rail in collaboration with the University of Twente and the Saxion University of Applied Science began the digitalization of its resources as the necessity for such a process began emerging worldwide. The main focus in this process is the catenary arch as it plays a crucial role in the network’s fundamental functionality. Such an arch consists of poles, insulators and many more elements. Currently, the companies managing the network do not have an overview of their infrastructure or the inventory of the elements. Moreover, it can contain both legacy and new elements, which requires trained professionals to perform a manual on-site state assessment and therefore makes the operations cumbersome and time-consuming. This is why there is a need for digital twins (virtual representations) of those, allowing for faster and more reliable repairs and improvements. This paper explores the process of matching the LiDAR-produced point clouds of the catenary arches to its CAD template stored in the catalogue form. It omits the semantic segmentation in its core as it is perceived as unnecessary and focuses on a deep convolutional neural network (CNN) to extract the features from a scan. Such an approach ensures the labelling task and retraining of the semantic segmentation network are unnecessary. This research is a proof of concept and confirms that the proposed method is feasible and provides numerous improvements over the existing one. To find the best interest point detection method for the given dataset, multiple methods were compared. In the last phase, it was compared to the existing method and provided an overall 250 times speed improvement in terms of adding a new CAD template to the models' catalogue.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Computer Science BSc (56964)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page