[GRASS-dev] New georectifying module in TclTk
Glynn Clements
glynn at gclements.plus.com
Tue Jun 6 17:01:20 EDT 2006
Michael Barton wrote:
> Thanks. That is helpful. It will be certainly easier if I only am doing rms
> calculations using one equation. Still, I wonder can minimizing errors using
> an affine transformation produce poorer results for polynomial
> transformations?
If you intend to use a polynomial transformation, the error
measurements for an affine transformation aren't really relevant.
In any case, the error measurement is only really useful to
Note that the fitting and measurement performed by i.points has no
effect upon the transformation chosen by i.rectify. It's just giving
the user an idea of what to expect from i.rectify (assuming that an
affine transformation is used), as well as highlighting points which
the user might have misplaced.
> Also, in responding to my own comment yesterday... I looked at gdalwarp last
> night and I'm not sure it could replace i.rectify as is. If there is someway
> to specify a set of GCP's that are not embedded in the raster file format, I
> didn't see it in the docs. Also, I didn't see any way of using it to do
> error calculations (though this might be moot). GDAL will also need to to
> GRASS to GRASS file formats. I know this was a problem awhile back, but
> perhaps it is not one now.
It would be useful if the same front-end could be used for both
i.rectify and gdalwarp. The main issue is that you need a way to
compute the best-fit transformation from the control points and to
project sample points without actually transformating the entire
image.
> More interestingly, it might be more possible to wrap v.transform into the
> georectifier so that it will do both raster and vector files, if this is of
> any interest to people.
v.transform also uses an affine transform. Non-affine transformations
on vectors are awkward as they transform straight line segments to
curve segments. Also, vector transformations are normally done in the
opposite direction to raster transformations.
--
Glynn Clements <glynn at gclements.plus.com>
More information about the grass-dev
mailing list