[Liblas-devel] pondering about LAS 2.0 (was LAS Schema)

[Martin Isenburg]

hello howard,

i agree with everything howard says. and i think that we can remedy all 
those things with LAS 2.0. i got the following idea from talking to 
someone at the IMLF showfloor (it was not my idea ... i stole it but i 
personally think it's brilliant at the moment but you may convince me 
otherwise):

make each point attribute a separate file. hence a "LAS file" is really 
a "LAS directory". then the presence of the file with the right name 
(e.g. "x", "y", "z", "rgb" ... determines the composition of the points 
(rather than a point format in the header).

C:\released_code\lastools\data>dir serpent_mount_las
 Directory of C:\released_code\lastools\serpent_mount_las

02/04/2010  11:24 PM    <DIR>          .
02/04/2010  11:24 PM    <DIR>          ..
01/27/2009  11:55 PM         227 header
01/27/2009  11:55 PM         5,966,832 x
01/27/2009  11:55 PM         5,966,832 y
01/27/2009  11:55 PM         5,966,833 z
01/27/2009  11:55 PM         2,983,416 r
01/27/2009  11:55 PM         2,983,416 g
01/27/2009  11:55 PM         2,983,416 b
               7 File(s)
               2 Dir(s)   3,245,469,696 bytes free

this takes away all the fluff. makes it super extendable. allow for more 
complex data types to be easily added (e.g. doubles, velocity vectors, ...).

and when you "classify" the LIDAR points you can just create a 
"classification" file without having to touch gigabytes of data just to 
re-write the classification entry of every point.

of course ... the disadvantage will be the multiple files overhead. but 
assuming we have spatial indexing as well we can get away with one very 
LARGE las files instead of tons of smaller las files, thereby lowering 
the total number of files again,

what do you think? feel free to set me straight now ... it was not my 
idea in the first place so you can't hurt my feelings. (-:

martin

-- 
scientist in data visualization  facilitator for sustainable communities
http://downtownfarm.net   925-456-4461  http://www.cs.unc.edu/~isenburg/