Powerline extraction from aerial and mobile LiDAR data using deep learning

Accurate powerline classification from LiDAR point clouds is essential for efficient monitoring and management of power distribution networks. Currently, the classification is being done through manual labeling or some semi-automatic methods which are time-consuming and resource intensive. In this study, we explore three deep learning architectures, namely KPConv, PointCNN, and RandLA-Net, for powerline segmentation in aerial and mobile LiDAR datasets. We utilize manually labeled aerial and mobile LiDAR datasets from Surry in Canada and Kerala in India, respectively. Impressive results are obtained, demonstrating high powerline classification accuracy and precision. KPConv outperformed the other architectures, achieving over 98