Autor: Elżbieta Zimoląg, B.Sc.
Titel: Detecting power lines and cables by TLS
Art: Master's Thesis
Jahr, Nr.: 2014
Betreuer: M. Sc. Wolfgang Wiedemann

Currently used techniques of overhead cable measurement (OCM) using tachymeters is a time consuming and expensive task. Terrestrial laser scanning (TLS) offers short measuring times, but manual post-processing is required to find points of interest (POI). Automatic POI extraction can help operators and improve the efficiency. This Master’s Thesis presents a procedure which can automate POI extraction from a dataset, that has previously been cleaned of irrelevant features. It focuses on representing the overhead cable grid with a minimal amount of parameters, while achieving the best possible precision. Current procedures for OCM are compared and and TLS is examined before the problem of 3D multi model recognition is transformed into a local peak search in two dimensions using the Hough transformation, which is theoretically examined and implemented iteratively in MATLAB with descriptions and illustrations. From the two dimensional feature extraction points on the overhead cables are extrapolated and parameters for catenary line functions for these cables in 3D are constructed using non-linear regression. The proposed algorithm is tested on two sets containing ~700 k and ~ 16 M points respectively and compared to the results of manual POI extraction from tachymetry data. As a result, a procedure was created to extract POI, and as a result yields the parameters for a set of catenary line functions in three dimensional space, representing the input grid with high precision for the test data. It is recommended the proposed algorithm be implemented in C++ with the OpenCV framework to optimize runtimes for maximum efficiency. This Master’s Thesis shows that automatic POI is a feasible solution for tasks requiring the determination of cables’ positions with an accuracy of about 1.4 cm, and lower accuracy on dense cable intersections.