An Interactive Visualization System for Huge Architectural Laser Scans

This paper describes a system for rendering large (billions of points) point clouds using a strict level-ofdetail criterion for managing the data out of core. The system is comprised of an in-core data structure for managing the coarse hierarchy, an out-of-core structure for managing the actual data and a multithreaded rendering framework that handles the structure and is responsible for data caching, LOD-calculations, culling, and rendering. We demonstrate the performance of our approach with two real-world datasets (a 1.8 b points outdoor scene and a 360 m points indoor scene).