Vario-scale visualization of the AHN2 point cloud

Abstract

LiDAR technologies are used to measure point cloud data of the earth's surface. The usage of LiDAR allows for the fast collections of massive data sets. The AHN2 point cloud data set, part of Rijkswaterstaats initiative to map the surface of the Netherlands, contains 639 478 217 460 points. For efficient visualization in web viewers, these massive point clouds are stored in an octree data structure. Visualization through this method has the downside of discretely visualizing the point cloud. These discrete artefacts are referred to as density jumps, and are visible where there is a boundary between blocks retrieved from the octree. These blocks contain different densities because they are retrieved from different levels of the octree. This thesis proposes a continuous visualization method for massive point cloud data sets that aims to eliminate these density jumps. While the continuous visualization of vector data sets has been extensively researched, this is a novel field of research for point cloud data sets. This thesis explores the feasibility of a vario-scale visualization method, and aims to implement it in an existing web viewer architecture. Due to the massive nature of the AHN2 data set, cloud computing and distributed computing techniques are used to imrove the workflow. The presented methodology removes the \textit{density jumps} by determining an upper density bound for the point cloud density relative to the camera position. Circle packing theory is used to reinforce the upper bound continuously, thus removing artefacts created by discrete density jumps. A proof-of-concept for this theory is implemented in an existing point cloud web viewer architecture.