Quantized mesh format: how accurate the tile s bSphere needs to be


I have a simple conceputal question. In your description of the quantized mesh format, one needs to know the bounding sphere of a tile. After a little research, I realize that algorithms calculating exact bounding spheres are quite complex and demanding.

I found much simpler algorithms but their accuracy is variable. Between 5 and 20% bigger than the optimum. I realize that all that might be resolution dependent, but how accurate do we need to be?

Thank you for your anwsers.

Kind Regards

The bounding sphere you supply must contain all the vertices, but being bigger than optimal won’t break anything, other than rendering some extra tiles unnecessarily. Cesium has code to compute a bounding sphere from vertices in the BoundingSphere class.


Hi thanks fo the reply, all I needed to know for now.

Hi, I’ve implemented a tool to convert folders of ASC and other DTMs to quantized-mesh format.
I’ve used Miniball to calculate the minimum bounding sphere (http://software.icecube.wisc.edu/icerec_trunk/doxygen/phys-services/Miniball_8h_source.html)

The results obtained are more than satisfactory.

Here is a screenshot: http://www.daigio.it/qmeshsample.png