[gdal-dev] RPC Transform Error: ERROR 1: latitude or longitude exceeded limits

[Claire Porter]

Yes, I am sure I entered the -te parameters correctly.  When I also
use -tr to specify the output pixel size, I get the correct result in
this one test case.  Is there no other option than this work around?
It seems like some sort of systematic issue with RPCs that would be
good to address at the source.

I will test some more and let you know if the work around fails in any cases.

Thanks,
Claire


On Mon, Apr 9, 2012 at 1:02 PM, Etienne Tourigny
<etourigny.dev at gmail.com> wrote:
> The error you get can usually be avoided when you set the output
> bounds correctly. Are you sure the -te options you use are correct?
>
> ERROR 1: Too many points (441 out of 441) failed to transform, unable
> to compute output bounds
>
>
>
> On Mon, Apr 9, 2012 at 1:57 PM, Claire Porter <claire.porter at gmail.com> wrote:
>> Etienne,
>>
>> When I set the extents using -te, I get an output of size 0.  I think
>> the problem is not in the warping itself, but in the transformation
>> using the RPCs.
>>
>> Thanks,
>> Claire
>>
>>
>> On Mon, Apr 9, 2012 at 10:04 AM, Etienne Tourigny
>> <etourigny.dev at gmail.com> wrote:
>>> Claire,
>>>
>>> You may have to set the output extents manually with the -te option.
>>>
>>> This has been the solution to many problems with gdalwarp.
>>>
>>> Etienne
>>>
>>>
>>> On Mon, Apr 9, 2012 at 11:12 AM, Claire Porter <claire.porter at gmail.com> wrote:
>>>> Dmitry (and the list),
>>>>
>>>> Using the warp options you suggested does not make a difference.  Even
>>>> SAMPLE_GRID does not significantly change the output.  I also tried using
>>>> wgs84 as the output coordinate system in case it was an issue of being
>>>> projected into Polar Stereographic: no luck.
>>>>
>>>> I copied the relevant functions from gdal_rpc.cpp into python order to test
>>>> the RPC calculations independant of the warping functions and any projection
>>>> system.  I used a sampling grid over the whole extent of whatever test image
>>>> and found that the RPCs do not converge for any part of the error-prone
>>>> images  In fact, with each iteration the offset from the original pixel gets
>>>> further and further from zero (alternating sign each iteration).
>>>>
>>>> From these tests, I am guessing that the issue is not the sampling or the
>>>> projection system.  Since other software packages can handle these images, I
>>>> assume then that the issue is in how gdal is managing the RPCs.  Any ideas?
>>>>
>>>> Many thanks,
>>>> Claire
>>>>
>>>> _______________________________________________
>>>> gdal-dev mailing list
>>>> gdal-dev at lists.osgeo.org
>>>> http://lists.osgeo.org/mailman/listinfo/gdal-dev