Transmission Tower Classification Using Point Cloud Similarity

Right-of-Way managers have increasingly used LiDAR inspections as an input to monitoring and maintenance activities of their infrastructures, making up a large percentage of the volume of data stored. Much of the shortcomings of this use revolve around the ability to accurately process data, classify elements and apply fitting monitoring strategies. This issue is raised by TSOs, when linking overhead line transmission tower scans to their respective models. In this sense, this work proposes a similarity based classification methodology to perform this task, supported by traditional point cloud distance metrics, using a set of Base Reference Models (BRM) - models built on alignment algorithms applied to pre-existent point clouds. This work tests this methodology for different sets of BRMs and point cloud distance metrics. We find that the effectiveness of this approach is highly related to the BRM resolution and to the distance metrics employed. For the use case at hand, the Chamfer distance similarity approached reached an accuracy as high as 89%.