Exploring Point Cloud Interaction and AI Features for Sewer System 3D-Scans

Abstract. Visual analysis of point clouds from Laser Scanning of complex structures, especially those scanned from the interior, can be a complex, multi-step process due to intricate geometries, overlapping surfaces, and the relatively small size of internal components. Large underground sewerage systems are often surveyed using Laser Scanning to document a complete 3D scan of the interior. In such cases, visual analysis of the 3D point cloud can be very useful for detecting problems, planning repairs, and communicating information among planning, engineering, and maintenance teams. Due to the complex systems of pipes, structures, instruments, and rooms that are featured in such 3D scans, combinations of interaction features and interfaces are required for the best possible workflows. In this paper, we present the initial development of a Unity-based prototype system for visual analytics of pre-processed 3D scans utilizing a combination of widely used techniques like camera view bookmarks, point cloud class highlighting, and exploded views, together with modern Large Language Model (LLM) assisted visual analytics. The proposed prototype was demonstrated to two sewer utility companies, where prospective users provided a positive initial evaluation of the presented features, even with the added complexities of required pre-processing and an LLM model. The prototype will be used as a basis for the development of future systems. Prototype link - https://github.com/Tannez/PointCloudInteractiveViewer/