Cross-Source Registration of 3D Model and As-Built LiDAR Point Cloud for Electric Power Transmission Lines Model Verification

Abstract

Due to the reality gap between designed models and as-built real-world scenarios, improving the accuracy of the 3D models of power transmission lines is essential when transferring them from the design department to the operation department. Point clouds obtained from the drone's remote sensing provide a credible reference for the verification of the 3D model. However, the accuracy of power pylon localization using traditional point cloud segmentation methods is limited due to the complex background of power transmission lines as well as the noise in the point cloud. In this paper, we propose a sequential pylon pose verification method based on cross-source point cloud registration. The method utilizes the registration of pairwise pylon point cloud iteratively for computing each pylon's pose relative to the first pylon. The registration algorithm based on the truncated least squares cost function ensures that our method avoids the pylon localization and orientation errors caused by inaccurate pylon segmentation. Therefore, our method only requires coarse pylon segmentation in point clouds. Experimental results validate the effectiveness of the proposed method in terms of verifying the position and elevation of electric pylons with both identical shapes and shape variations between the source and target. © 2025 Institute of Electrical Engineers of Japan and Wiley Periodicals LLC.