If the point density is so high, then you almost certainly don't want
the extra overhead of having to process every single point 121 times!
I strongly suggest you instead look at downsampling the point cloud to a
density which is better suited to the size of the features you are
trying to extract.
For my orienteering maps I also find industrial buildings in some areas,
but I will much rather accept that they initially can end up in 2 or
more tiles.
Anyway, the idea behind buffered tile-based processing is to allow every
tile to be handled in isolation, on arbitrary many processing cores.
Terje
--
- <
Terje.M...@tmsw.no>
"almost all programming can be viewed as an exercise in caching"