read function for LAS data

Michael,

I have not gone through extensive testing, but it seems pretty fast and
usefull for my 1.0 las. I've loaded a 235 Mb file on a USB drive, and it ran
in ~20 seconds. SAGA, in comparison did it in 3 min 22 seconds.

Cheers,
Etienne

Le 2010-06-06 10:18, Michael Sumner a ?crit?:

Hello,


To other issue I see here is that R is loading the whole file in memory, so
if you can manage small files, that might not be that easy with (standard)
larger ones. Don't you think ?


That was certainly true of the version of code I posted, but writing a
more flexible version is not difficult, and actually less difficult
than I expected. I've implemented arguments to "skip" and read "nrows"
at a time, so there is the beginnings of a wrapper around the core
read for building more flexibility.
(I was thinking of including subsetting of various kinds which really
makes it more complicated, and the appropriate level to handle that is
in a wrapper to this function).
I've updated the R source on my site, and here's a new example. This
should be considered as a rough working draft, the details can be
hidden in the final suite of functions. My chunk/rows handling is
pretty awkward, and may have bugs for particular record numbers.
Any testing you can provide would be greatly appreciated.
# new version with "skip" and "nrows" arguments
source("http://staff.acecrc.org.au/~mdsumner/las/readLAS.R")
f <- "lfile.las"
## get just the header
hd <- readLAS(f, returnHeaderOnly = TRUE)
numrows <- hd$`Number of point records`
## [1] 1922632
## read in chunks, and pass to DB or ff, or subset by sampling, etc..
rowskip <- 0
chunk <- 1e5
rowsleft <- numrows
system.time({
## keep track of how many rows we skip, and how many are left
for (i in 1:ceiling(numrows / chunk)) {
	if (rowsleft < chunk) chunk <- rowsleft
	if (chunk < 1) break;
	d <- readLAS(f, skip = rowskip, nrows = chunk)
	rowskip <- rowskip + nn
	rowsleft <- numrows - nn
}
})
#   user  system elapsed
#   1.10    0.55    1.64
On Sun, Jun 6, 2010 at 8:09 PM, Etienne Bellemare Racine
<etiennebr at gmail.com> wrote:


This is interesting, I'll try your code on my lidar files in the next few
days.
?2010-06-04 22:36, Michael Sumner wrote :


Thanks Alex, I will eventually post this to a broader audience.
I've used liblas and lastools, but the aim here is for a pure R
implementation that is built directly from the LAS specification
without 3rd party tools.


What might be of interest in using liblas is that it provides support for
many las versions and they plan to provide support for some versions to come
(conditional to funding) so having an R binding might be of interest here.
They are also working on the integration of a spatial index which would
allow easier handling of large files. I must say I don't know how hard
writing a wrapper for R might be for that particular tool.
To other issue I see here is that R is loading the whole file in memory, so
if you can manage small files, that might not be that easy with (standard)
larger ones. Don't you think ? Did you give a try to the R SAGA package.
There is a module for loading las files but again, I don't know how it
manages memory. I guess that it could be possible to use some sort of ff
package to handle bigger files, but that's just on the top of my head.
Etienne


The R code already works quite well to extract x/y/z/time/intensity,
it just needs some extra work to tidy up and generalize things and
ensure that very big datasets can be read.
Cheers, Mike.
On Sat, Jun 5, 2010 at 6:07 AM, Alex Mandel<tech_dev at wildintellect.com>
?wrote:


On 06/03/2010 07:54 PM, Michael Sumner wrote:


Hello,
I'm looking for interest in this functionality and eventually working
it into a package.
I don't actually use LAS data much, and only have one example file so
I'm hoping others who do or know others who would be interested can
help. I have previously posted this to r-spatial-devel.


I think there are people who would use it. You might want to have a look
at http://liblas.org/ (some of the same people that do gdal/org work)
Wrapping this library might be a good approach. There are example files
available too.
Thanks,
Alex