[gdal-dev] Yet another GML EPSG::4326 offsetvector XY-order problem

Riivo Kolka riivo.kolka at gmail.com
Thu Mar 20 02:05:30 PDT 2025


GDAL_JP2K_ALT_OFFSETVECTOR_ORDER=YES helps, thanks.

The spec, though. Holy Space Crab! Verbosity over clarity.

Kontakt Even Rouault (<even.rouault at spatialys.com>) kirjutas kuupäeval
K, 19. märts 2025 kell 20:52:
>
> Hi,
>
> This is a recurring source of confusion as the related specs aren't very explicit / clear about that. But all example in https://docs.ogc.org/is/08-085r8/08-085r8.html use
>
>         <gml:offsetVector>0 positive_value</gml:offsetVector>
>         <gml:offsetVector>negative_value 0</gml:offsetVector>
>
> for EPSG:4326, so at least the behavior of GDAL is consistent with that, and thus I'd tend to believe your Pleiades images are faulty
>
> Rather than hacking the file, you might try setting the environment variable / configuration option GDAL_JP2K_ALT_OFFSETVECTOR_ORDER to YES, although I'm not 100% positive it does the exact switch of coefficients that is needed here...
>
> Even
>
> Le 19/03/2025 à 18:51, Riivo Kolka via gdal-dev a écrit :
>
> Hi.
>
> I see that gdaljp2metadata.cpp has lots of code for extracting
> something useful but I got files that still escape the effort.
>
> Some Pleiades jp2 images that gdalinfo fails to find the
> Origin and Pixel Size and reports wrong Corner Coordinates.
> gdalinfo only reports (seemingly correct) GeoTransform
>
> I verified that just the order of    <gml:offsetVector/>s matter.
> If i switch the values inside jp2 <gml:offsetVector/> then gdal detects correct
> Origin, Pixel Size and correct Corner Coordinates.
>
> GDAL 3.10.2, released 2025/02/11
>
> Full description in a gist
> https://gist.github.com/rkolka/70ef4ac1172d6ed98b51d0412ff73d49
> _______________________________________________
> gdal-dev mailing list
> gdal-dev at lists.osgeo.org
> https://lists.osgeo.org/mailman/listinfo/gdal-dev
>
> --
> http://www.spatialys.com
> My software is free, but my time generally not.


More information about the gdal-dev mailing list