Clarifying geometric error in Cesium World Terrain quantized mesh tiles

Cesium ion
1

Hi all,

I’m working on a flight path analysis system that uses terrain data for clearance calculations, and I have a few questions about how geometric error is defined and exposed in quantized mesh terrain tilesets.

Pipeline understanding

My current understanding of how a tileset like Cesium World Terrain is built is roughly: source DEMs are mosaicked and reprojected into a tiled raster, which is then triangulated into a TIN per tile,similar in spirit to the approach used by tin-terrain. In that model, the geometric error for a given tile would represent the maximum vertical deviation between the TIN mesh and the source raster.

Is that an accurate high-level description of the pipeline? Or is the geometric error defined differently, for example, as a function of the source data resolution rather than the actual triangulation error?

Concrete example — Cesium World Terrain in Canada

The Cesium World Terrain data sources page lists an approximate resolution of 30 m for Canada. My question is: does the geometricError in the tileset JSON reflect the actual mesh-vs-raster deviation for those tiles, or is it derived from the nominal source resolution (e.g., 30 m projected to screen-space error)?

Per-tile geometric error

Finally, is per-tile geometric error available anywhere, either in the tile metadata or via a separate endpoint? I’m interested in choosing zoom level based on geometric error.

Thanks in advance, happy to provide more context about the use case if helpful.

2

Hi Alexandre,

For quantized mesh, you can find the approach for screen space error here → Hausdorff distance - Wikipedia . Per-til geometric error is available with 3D Tiles, but not with quantized mesh.

For your question about geometricError in Canada at 30m resolution, the resolution represents the highest LOD available for that dataset, are you wondering how that computes to SSE?