Hi All,
I'm looking for some ideas on how to get the z value from a point cloud onto a point in porformant way.
As far as I can tell there is no easy way to do it. Am i missing something?
Options I can see available:
Convert to Raster with RasterDEMGenerator and use PointOnRasterValueExtractor - pretty cpu and memory intensive
Convert to points and use NeighborFinder - pretty cpu and memory intensive
Has anyone got this process figured out?
What I did was buffer the requestpoints, add unique id's , clip pointcloud with buffers (merge unique id's), convert clipped pieces to points then neighborfinder (group by unique id). This worked ok-ish for my volume of data. But would love to hear alternative solutions as well.
I think having the pointcloud in a database and being able to select a subset based on location, not having to load all the data every iteration, would be a good start.
What I did was buffer the requestpoints, add unique id's , clip pointcloud with buffers (merge unique id's), convert clipped pieces to points then neighborfinder (group by unique id). This worked ok-ish for my volume of data. But would love to hear alternative solutions as well.
I think having the pointcloud in a database and being able to select a subset based on location, not having to load all the data every iteration, would be a good start.
Yep, that was a good suggestion. When using clippers first there was barley any memory overhead. I will see if I can try and get some parallel performance going where possible. Will report back if successful.
Time is definitely taken in the clipping so perhaps I can split the load
Yep, that was a good suggestion. When using clippers first there was barley any memory overhead. I will see if I can try and get some parallel performance going where possible. Will report back if successful.
Time is definitely taken in the clipping so perhaps I can split the load
Sticking part of the process in a CutomTransfomer and processing each tile in parallel sped things up significantly. I put the FeatureReader along with the clipper inside and PointCloudCoercer. Few features in, few features out. NeigborFinder was used outside just so that all points could be considered from all tiles.