[gdal-dev] GCP list projection

[Frank Warmerdam]

On 11-02-17 10:08 AM, Vadim Shlyakhov wrote:
> Hello,
>
> I wonder if it makes any sense if GCP list at a dataset is different from the
> dataset's SRS.
>
> I've played a little with these in Python, but neither gdal.Transformer, nor
> warping via VRT doesn't produce the results expected. I mean, presumably if
> GCP's SRS is different from the dataset's SRS, then they are to be transformed
> internally into the dataset SRS, but this doesn't seem to happen

Vadim,

Normally I would expected a dataset with GCPs to have a GCP SRS, and no
dataset level projection or geotransform.

When given such a datsaet, gdalwarp will attempt to produce an output
dataset in the *GCP SRS*.

If the source dataset has both a GCP SRS and a dataset level SRS I am not
absolutely positive what gdalwarp does, but I feel it *ought* to be producing
an output file in the GCP SRS if it is using the GCPs.  I am not sure why you
expect the GCPs to be reprojected.

Note, if the output file is in a different SRS than the GCPs, I think the GCPs
would still be used to compute a polynomial in the SRS of the GCPs for
getting from source file pixel/line to source georeferenced coordinates.  Then
reprojection would be applied as an extra step.

In theory it could make sense to have a dataset with a dataset level SRS
and geotransform, and also GCPs with a different SRS.  In some cases the GCPs
are in a particular SRS (say lat/long) just because it is a convenient
coordinate system to collect the GCPs, not because it is representative of
the geometry of the image.

Best regards,
-- 
---------------------------------------+--------------------------------------
I set the clouds in motion - turn up   | Frank Warmerdam, warmerdam at pobox.com
light and sound - activate the windows | http://pobox.com/~warmerdam
and watch the world go round - Rush    | Geospatial Programmer for Rent