A novel method to segment individual wire from bundle conductor using UAV-LiDAR point cloud data. (April 2023)

• Record Type: Journal Article
• Authors: Shen, Yueqian; Yang, Ye; Jiang, Jundi; Wang, Jinguo; Huang, Junjun; Ferreira, Vagner; Chen, Yanming
• Is Part Of: Measurement. Volume 211(2023)
• Publication Date: 2023-04
• Subjects: Powerlines -- UAV-LiDAR -- Point cloud -- CFDP algorithm -- Wire segmentation; Weights and measures -- Periodicals; Measurement -- Periodicals; Measurement; Weights and measures; Periodicals; 530.8
• DOI: 10.1016/j.measurement.2023.112603
• Languages: English
• ISSNs: 0263-2241

Abstract

Highlights: Slicing idea combined with CFDP algorithm is proposed to segment wires from bundle conductors. To eliminate the influence caused by density variation, the model for parameters setting against density variation is established. Abstract: Monitoring and managing the powerline corridors are essential in the electricity demand of our daily activities and productions. UAV-LiDAR is being used as a feasible technique for such tasks to reduce the amount of manual inspection of the powerline corridors. However, challenges in extracting powerlines using a large point cloud remain due to the various scenarios and unavoidable noise. In this work, a novel procedure to segment wires individually from bundle conductors automatically using the point cloud acquired by the UAV system is proposed. The scene is first voxelized, and the voxel-based heigh features are generated and utilized to rough detect the locations of the objects. The powerline span is determined using the adjacent pylons, and the corresponding powerlines are extracted. The powerlines are segmented to different bundle conductors using the connected-component analysis in one powerline span. Finally, the slicing procedure is implemented along the powerline span direction, and CFDP (clustering by fast search and find density peaks) algorithm is introduced to segment the individual wires from the entire bundle conductors. To solve the failure results caused by various density, the parameters used in the CFDP algorithm is optimized. This procedure was validated by comparing powerlines extracted manually using the CloudCompare software. For the uniform density bundle conductors, the quantitative assessment in terms of precision, recall, and F1-score are 98.17%, 96.60%, and 97.37%, respectively. The corresponding values for the nonuniform density bundle conductors are 95.25%, 88.03%, and 90.95%, respectively. Results of the proposed method are compared to k-means, DBSCAN, and spectral clustering, demonstrating the superiority of the effectiveness.