Skip to main content Skip to main navigation

Publication

Computing Arbitrarily Large Meshes with Level-of-Detail Support for Cesium 3D Tiles

Malte Hillmann; Felix Igelbrink; Thomas Wiemann
In: Proceedings of the 7th International Workshop on LowCost 3D. LowCost 3D - Sensors, Algorithms, Applications (LC3D-2022), December 15-16, Würzburg, Germany, ISPRS Archives, 2022.

Abstract

3D data of large scale environments are usually too large to be rendered on consumer hardware, although many approaches to reduce the amount of data exist. One example are octrees, that are used to subsample point cloud data. Over the last years, the Cesium engine has proven to be a good tool to stream geospatial data for rendering on low-cost hardware. Newer versions of the used Cesium 3D Tiles format also include support for 3D meshes. However, such meshes are seldom used in the context of rendering of large scale data. In this paper, we present an approach to generate large scale meshes for Cesium 3D Tiles. The main advantage of generating such meshes over point clouds is that the amount of data transferred is significantly reduced. Our method is able to generate the required data to stream 3D meshes in the Cesium file format directly on consumer-grade hardware with a constant memory footprint, allowing us to generate surface reconstructions from arbitrarily large point clouds. Using the freely available web viewer offered by Cesium, the generated meshes can be deployed via a web server and rendered in a browser with the Cesium Viewer. All presented algorithms are freely available under the BSD 3-clause licence.