[Liblas-devel] Open source lidar software?

Hamish hamish_b at yahoo.com
Thu Apr 7 04:37:11 EDT 2011


francesco wrote:
> GRASS is great up the the million points limit, (2 Gb machine)
> then things get problematic

hint:
when importing as vector points with v.in.ascii you need to
use the -b flag to avoid building topology. If you do that you
can import as many points as you like without the RAM overhead.
..There is a small but finite memory overhead req'd for every
feature with topology. After 2-3 million points that becomes an
issue on a desktop with 2GB RAM.

Also by default a simple x,y,z[,category integer] import will
not create an attribute database. Add more (non-spatial) columns
and by default it will create a DB. A database associated to your
data via the category index is obviously some overhead too..
You can avoid building an associated attribute database with the
v.in.ascii -t flag. This does mean unfortunately that if you have
x,y,first hit,last hit,return strength,... data you'd have to
import it as multiple maps, assigning the "z" to other meanings.
On the plus side you can then use the d.vect module's -z flag
to instantly colorize the rendering using any of the assortment
of standard or custom color rules.
You won't be able to do fancy stuff like network analysis without
topology, but who uses that for points?


when importing with r.in.xyz & piping from stdin, there is no
limit to the size of import. (except for advanced methods like
trimmed mean which eats RAM; not sure about skewness)
When importing from a file you're limited by your OS/filesystem/
glibc. actually a 32 bit glibc might just break the percentage
done status [ftell()], but the import would still work fine.
And if you're starting from a file with random access ability you
can switch on multi-pass import mode which means the RAM
consuming trimmed mean (etc) can run in multiple passes,
effectively avoiding the RAM problem.


> Postgres/postgis is great for playing with complex queries but
> what about visualization?

no idea if they do, but if .vtk viewers like Paraview or VisIt
had direct PostGIS support it would mean all sorts of awesomeness.
  http://www.paraview.org/paraview/project/imagegallery.php
  https://wci.llnl.gov/codes/visit/gallery.html


regards,
Hamish



More information about the Liblas-devel mailing list