[gdal-dev] Problem reading Sentinel-1 GeoTIFFs

[Simon Eves]

Further to my earlier post about GCP Transforms, I now have that code
written, and it seems to be functioning fine, although whether it's
generating the right values is as yet unclear.

The problem is that I am having trouble creating the secondary
transformation from whatever the output space of the GCP transform is, to
my desired space (4326 in this case).

Whenever I call GetGCPSpatialRef() on the datasource, I get the following
error (twice)

  PROJ: proj_create_from_database: ellipsoid not found (1)

I get the same error (also twice) when I run gdalinfo on the file, but then
with a plausible looking SR dump, which may of course just be a default:
______________________________________________________

$ gdalinfo
s1b-iw-grd-vv-20211110t033513-20211110t033538-029520-0385e8-001.tiff
ERROR 1: PROJ: proj_create_from_database: ellipsoid not found
ERROR 1: PROJ: proj_create_from_database: ellipsoid not found
Driver: GTiff/GeoTIFF
Files: s1b-iw-grd-vv-20211110t033513-20211110t033538-029520-0385e8-001.tiff
Size is 25541, 16650
GCP Projection =
GEOGCRS["WGS 84",
    DATUM["World Geodetic System 1984",
        ELLIPSOID["unnamed",6378137,298.25722356049,
            LENGTHUNIT["metre",1]]],
    PRIMEM["Greenwich",0,
        ANGLEUNIT["degree",0.0174532925199433]],
    CS[ellipsoidal,2],
        AXIS["geodetic latitude (Lat)",north,
            ORDER[1],
            ANGLEUNIT["degree",0.0174532925199433]],
        AXIS["geodetic longitude (Lon)",east,
            ORDER[2],
            ANGLEUNIT["degree",0.0174532925199433]],
    ID["EPSG",4326]]
Data axis to CRS axis mapping: 2,1
GCP[  0]: Id=1, Info=
          (0,0) -> (45.0323375081341,62.6475859336067,149.992539359257)
[etc.]
______________________________________________________

My constructed OGRCoordinateTransformation for the second transform is
therefore a no-op, but the final values (which are therefore just the
values out of the GCP transform) are definitely not 4326.
______________________________________________________

omnisql> select * from s1 where rowid<10;
raster_lon|raster_lat
476788.1263799404|747980.9704571336
460813.8583986053|747120.1413700345
444852.7791957706|746184.4282221086
428904.8887714364|745173.8310133558
412970.1871256025|744088.349743776
397048.6742582691|742927.9844133693
381140.3501694361|741692.7350221358
365245.2148591036|740382.6015700753
349363.2683272713|738997.5840571877
333494.5105739397|737537.6824834733
______________________________________________________

Those are the first ten pixels of the first scanline of the (~25K x ~16K)
file, btw, so intuitively there is something worrying me about those values
anyway (they are moving very quickly and non-linearly considering that's
just ten pixels).

Anyway, some Googling led me to...

  https://gitlab.orfeo-toolbox.org/s1-tiling/s1tiling/-/issues/46

...which seems very related, and even mentions a patch to libGeoTIFF by
Even. I checked, and that patch IS in the (internal, as configured in the
build) GeoTIFF in our GDAL.

This is with GDAL 3.4.0 and Proj 7.2.1.

Thanks in advance,

Simon

-- 
<http://www.omnisci.com/>
Simon Eves
Senior Graphics Engineer, Rendering Group
100 Montgomery St (5th Floor), San Francisco, CA 94104, USA


Email: simon.eves at omnisci.com | Cell:  +1 (415) 902-1996
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.osgeo.org/pipermail/gdal-dev/attachments/20211210/a4ae85eb/attachment-0001.html>