[Liblas-devel] estimating lines and samples

[Hamish]

> >> I'm writing a plugin for ossim to form an image
> from las files.  This will be open sourced as soon as
> it's useful.  Does anyone have any advise on how to
> choose a grid resolution which will determine the lines and
> samples?  I thought about using the square root of the
> number of points for a return and then scaling by ratio of
> the bounding box.  It would be nice to pick a grid size
> that gets at least one point for grid.  Any ideas?


it will be hard to beat GRASS GIS's r.in.xyz module!

As the author of it I'm totally biased of course, but it's true.
:-)

the only thing it doesn't do is multi-threading, but as it's
I/O limited (and blindingly fast) that's not really an issue.
Input filesize is only limited by operating/file system, or
no limit if piping from stdin. tested with 300+ GB input file
by the US Fish & Wildlife Service, only took it some hours to
churn through.

Massimo Di Stefano should be able to give you some tips on
smooth ossim+grass workflow/cooperation.


see http://grass.osgeo.org/wiki/LIDAR for tips on first creating
a "n" grid map, and then after clearing out the n=0 cells, using
univariate statistics on the rest to decide on your grid
resolution/density.

I'm stongly in favour of forcing the user to set the grid size
manually [ie thinking about it], but then I'm more concerned
about the data product than fast visuals.


regards,
Hamish