[Liblas-devel] LAS file Coordinate Systems
Howard Butler
hobu.inc at gmail.com
Tue Jul 21 21:17:15 EDT 2009
On Jul 21, 2009, at 6:46 PM, Mateusz Loskot wrote:
> Patel, Ronak Avinash \(US SSA\) wrote:
>> Mateusz,
>>
>> So you're saying the Geotiff specification describes how to handle
>> the
>> case where files cross UTM boundaries?
>
UTM boundaries are not hard. You can have a point that is "in" UTM
Zone 14 but its data are described in UTM Zone 15 and it is still
valid (though not as accurate as if it were actually projected into
Zone 14). To answer your question, however, neither the GeoTIFF
specification nor the LAS specification say anything about spanning
UTM zones within the same file. What is usually done is a coordinate
system that is local enough to the area of interest is chosen with a
specific scale and offset. For small areas, UTM is ok, but for
something like the state of Texas, you would want something else
(multiple UTM zones there).
By appropriately setting the offset and scale, I think you should be
able to finagle enough precision to work with WGS84 in LAS. Look at
some samples in the sample library <http://liblas.org/samples> and see
how they are setting scale/offset values for some ideas.
I would note that other than Mateusz and myself comfort and
familiarity with C/C++, one of the major reasons libLAS is implemented
in C++ is we can easily stand on the shoulder of other giants
(libgeotiff, GDAL, libspatialindex, etc). The Java world lags a bit
in the closer-to-the-metal geospatial libraries area.
I would suggest that you make a patch for GeoTools that is going to
allow you to get/set GeoTIFF keys and separate that from the GeoTIFF
header writing code. Otherwise, you are going to have to implement
all of the GeoTIFF "spec". I use spec in scare quotes because out in
the wild, it is much more widely (and poorly) implemented than even
LAS. Both GeoTools and libgeotiff (to an even stronger degree) have
smoothed over some of these inconsistencies, though thar be many
dragons there.
Howard
More information about the Liblas-devel
mailing list