A Robust and Efficient Method for Power Lines Extraction from Mobile LiDAR Point Clouds

Abstract

Monitoring, maintaining, and organizing power lines corridors are of great importance because they are a primary means to transfer generated electricity from power stations to surrounding areas. Mobile Terrestrial Laser Scanning (MTLS) systems have significant potential for efficiently creating a power line infrastructure inventory. In this paper, a novel algorithm is presented for automatically extracting utility poles and cables from MTLS point clouds in three consecutive phases of pre-processing, poles extraction, and cables extraction. In the pre-processing step, after dividing the MTLS data into several tiles or sections along the road and using trajectory data, noisy points and low-height points are eliminated from each section. Next, search areas containing lines are detected using a Hough Transform (HT) algorithm, and utility poles are identified based on horizontal and vertical density information. The search area for cables is estimated using a two-dimensional (2D) Delaunay Triangulation (DT) of the center points of the extracted poles as vertices. In each search area, high-density points are removed as non-cable points and utility cables are eventually extracted by fitting cable points to polynomial equations. The algorithm was tested on three different MTLS point clouds from a 1371 m urban road section, and a 2800 m and a 500 m non-urban road sections. Each of these datasets has unique challenges and was used to evaluate the efficiency of the proposed algorithm under different conditions. The algorithm was able to extract poles with average correctness of 100% (no false positives) and completeness of 97%. Similarly, average correctness and completeness of 100% and 95.6% were attained for cables, respectively. These detection levels show that the proposed method for power lines extraction from an MTLS point cloud is both reliable and feasible.