08/12/2026
Tech tip! Want to become a ground classification expert in SpatialExplorer?
The key to understanding the ground classification algorithm is that it uses a series of progressively finer triangulated surfaces created from low points. Each surface can only be made from points within a certain tolerance of the previous surface, which prevents the ground class from climbing vegetation or walls.
The first surface is made by sampling low points at a very coarse interval, typically 20 m. This coarse distance corresponds to the Cell Size parameter. Why would you sample at such a coarse interval? Building removal. If you find the lowest point every 20 meters, it’s very unlikely to be a building point.
After this first iteration, the next iteration samples low points every 2 meters, or whatever the Feature Size parameter is set to. This parameter is then decreased every successive iteration, so typically by the 10th iteration, low point sampling happens at a very fine grid size.
Urban data sets require a Cell Size of ~20 m for building removal. With forestry or non-urban data sets, particularly when there is significant terrain variation, a smaller initial Cell Size parameter will sample the terrain more precisely and result in a better ground classification.
