[gdal-dev] Thin Plate Spline

[Jan Hartmann]

At:

http://mapserver.sara.nl/downloads/gdal_tps.png

you can see a screenshot of the original image (right) and the 
georeferenced one (left, rotated almost 90 degrees). The red markers 
with the number 3 inside them show one of the control points: to the 
right the scan pixel  (5023/3421), to the left the targeted 
georeferenced coordinate in EPSG:28992 (121527/487174). As you see, the 
georeferenced point is not exactly on the border of parcel 7, as I 
expected. Above the map you see the four coordinates I used as control 
points; I added four extra points half way the original ones, so the 
complete warp was done with 8 control points.

Adding or removing points doesn't make any difference, and the other 
control points are shifted too, although slightly differently. When I 
use "-order 1" with the same control points I get exactly the same 
picture; however, with really weird control points the images for -tps 
and -order 1 get distorted in different ways.

It's only 10 meters or so, but it is very visible and too much for my 
purposes.  How can I get a georeferenced image with the control points 
exactly on their indicated location, no matter what happens in between?

Jan
>
> On 05/01/10 00:21, Even Rouault wrote:
>> It should work... Perhaps you can share how to reproduce what you see
>>
>> Le Saturday 01 May 2010 00:17:06 Jan Hartmann, vous avez écrit :
>>> Hi,
>>>
>>> I'm rubber-sheeting old, deformed maps using the thin plate spline
>>> option of gdalwarp (-tps). My impression was that the control points
>>> would be transformed to exactly their georeferenced locations, but in
>>> the georeferenced image the control poimts don't quite align with the
>>> gcp-coordinates I specified. I'm only using five points per map, so
>>> could this be the problem? If so, how many points do I need to get the
>>> control points on their exact location? Or are there other ways to
>>> achieve this?
>>>
>>> Jan
>>> _______________________________________________
>>> gdal-dev mailing list
>>> gdal-dev at lists.osgeo.org
>>> http://lists.osgeo.org/mailman/listinfo/gdal-dev
>>
>>