Terrain Rendering for GIS applications

For terrain rendering applications that work on remote sensing data, e.g. GIS applications, it is useful to allow the terrain data to change dynamically. For example, elevation data from different sensors can be compared or combined, or data sets from different points in time can be compared to analyze changes. Furthermore, some elevation data sets might require processing at visualization time, e.g. to fill holes with data from other data sets, or to remove unreliable parts of the data.

Dynamically changing terrain data poses several challenges to terrain rendering. Continuous level of detail (CLOD) techniques cannot rely on precomputed information since the geometry for the current view of the scene is not known a-priori. Instead, an adaptive mesh has to be created and/or refined on the fly using only the dynamically generated elevation data and little additional information.

Furthermore, most measured terrain data sets are given relative to the reference ellipsoid, not relative to a plane, and they cannot be transformed to a simpler reference geometry without loss of information. Accurate rendering of such data thus needs to be based on the reference ellipsoid. This poses several challenges that are often neglected in the Computer Graphics domain.

Example video of zooming in on the area around Rome, with several overlayed remote sensing data sets and dynamic terrain data changes: normal mode and wireframe mode.

Free source code for accurate rendering of planetary-scale, dynamic terrain data is available at https://marlam.de/ecm.

This research project was partially funded by grant KO-2960-3/1,2 from the German Research Foundation (DFG).