[GRASS-dev] grass6.2?

[Glynn Clements]

Michael Barton wrote:

> 1. In order to calculate RMS error, it is necessary to transform the GCP's
> unprojected coordinates into projected coordinates, and calculate the
> diagonal distance between the transformed coordinates and the coordinates
> used to define the transformation. Apparently, there is already C-code to do
> this in the imagery library. Glynn said that he might be able to do a quick
> C-module to accomplish this, using existing code rather than have anyone
> write new code to do it.

To clarify: the code to choose a transformation and to project
coordinates is in the imagery library, while the error calculation is
in i.points.

However:

1. i.rectify has its own alternatives to the functions in the imagery
library. The i.rectify versions support 2nd- and 3rd- order
transformations, while those in the imagery library only support
1st-order (affine) transformations. Thus the error calculations are
only meaningful if you use a 1st-order transformation with i.rectify.

2. GDAL contains its own version of the code used by i.rectify. Rather
than duplicate this, we should use the GDAL version.

> 2. I've combined raster and vector georectification into one module.
> However, the set up for georectifying vectors differs from rasters in
> several ways. Hamish and others suggested that this is a good time to make
> them consistent. This involves modifying v.transform and either extending
> i.group or making v.group. Here are the differences...

One difference which should probably be stated more explicitly is that
i.rectify can use a source from a different location, whereas
v.transform requires the source to be in the current location.

-- 
Glynn Clements <glynn at gclements.plus.com>