An Automated Method for Power Lines Extraction from LiDAR Data in a Complex Environment

Authors: HAITAO LYU*, University of Texas - Dallas, FANG QIU, University of Texas at Dallas
Topics: UAS / UAV, Mountain Environments, Geographic Information Science and Systems
Keywords: LiDAR, Power Line, Hough Transform,
Session Type: Virtual Paper
Presentation File: No File Uploaded


Extracting power lines in a complex environment, especially when transmission lines (115 KV to 230 KV ) and distribution lines (below 110 KV) coexist and are catenary and are not parallel to the ground, is still a huge challenges. To address this challenge, based on Hough Transform, Cylinder Fitting, Principal Component Analysis (PCA) and RANSAC (Random Sample consensus), an automated power lines extraction method in a complex environment was proposed in this paper. The method utilize a quad-tree to construct a local point cloud for each point and its n nearest points, and then a Cylinder Fitting model based on PCA was proposed to extract the features for each point from its point cloud, which can be used to evaluate a point the linearity, sphericity and planarity. Then The features were used to train a Gaussian Mixture Model to filter non-line points, and then a point cloud rasterization algorithm was put forward to transform residual points to a raster image. Finally, a power line corridor model based on Hough Transform and RANSAC was proposed to construct corridors based on the lines in the raster image, which can be used to locate the accurate positions of power lines. The related experiment results can prove that the method can extract more power lines and more accurate positions than other methods published in recent years.

To access contact information login