My first attempt was to download a raster TIFF image from http://esrinl-content.maps.arcgis.com/apps/Embed/index.html?appid=a3dfa5a818174aa787392e461c80f781. Unfortunately, when opening the 500MB images, most points where either #000000 or #FFFFFF, so I could not really use that. There are excellent viewers, along with height maps like https://ahn.arcgisonline.nl/ahnviewer/, so perhaps I just could not find a good source.
My second attempt was to visit http://ahn2.pointclouds.nl/. This data is slightly older (AHN2 from 2008 instead of AHN3 from 2015), but has similar high resolution (0.5 meter), and served my purpose. The websites has a great export function to select and download the region I was interested in. I downloaded two LAZ files with a 4mln and 150,000 point cloud, the later for testing.
I used the free and excellent CloudCompare with this workflow:
- Open the LAZ files
- Normalize the LAZ points. I used the settings "Quadratic" and "prefer +Z", because I read that the first is good for sharp edges on buildings, and the second good for terrain.
- Optionally subsample the point cloud to create a lower density point cloud (hoping that this would result in a smaller mesh in the next step)
- Use Poisson Surface Reconstruction to turn the point cloud into a mesh, as explained on the PSR wiki page.
- Export the Mesh in STL format (it also supports PLY, OBJ, VTK, OFF and DXF, but Sketchup comes with a free STL importer)
- Import the STL into Sketchup using the Sketchup STL importer.
- Wait till it is imported, and scale it up by a factor 1000x (the imported group was only a centimeter large in the model. I used the "Zoom to selection" function to find it :) )
This worked reasonably well for the sample export (170,000 points translated to 130,000 faces). It will not work for the 4,000,000 point export. Unfortunately, the subsampling resulted in a slightly larger mesh size (150,000 faces), not smaller, and I can't find a way for CloudCompare to reduce the complexity of the Mesh model. An alternative is to hand-pick points, and export that point list, but that seems rather tedious.