[GRASS-user] question on georectification

[Michael Barton]

The results are bad because each scan is slightly different. That is, you
probably had the map slightly rotated differently and placed differently on
the scanner in each scan. So you cannot use one set of ground control points
(GCP's) to rectify all scanned pieces.

The GCP's provide a mapping between the scanned image coordinates (xy) and
corresponding earth coordinates (east,north or long,lat). This mapping will
be different for each scanned part of the map. What are the earth
coordinates for x=100 y=100 in each image? They will be different for each
one.

You should be able to make the pieces match up pretty well, however, if you
make sure that some of your GCP's are in the parts of the scans that
overlap. 

Assuming that the original map that you scanned is georectified, you only
need a few points to perform a 1st order rectification (shift and rotate, no
warping). So it should go pretty fast.

You might want to try the new georectifier (under the file menu) now in the
cvs. IMHO, it makes this an easier job. If you use the georectifier, start
out in the location/mapset you want the scanned maps to be rectified INTO
rather than the location/mapset where the scans are now.

Michael
__________________________________________
Michael Barton, Professor of Anthropology
School of Human Evolution & Social Change
Center for Social Dynamics & Complexity
Arizona State University

phone: 480-965-6213
fax: 480-965-7671
www: http://www.public.asu.edu/~cmbarton



> From: Wolfgang <wollez at gmx.net>
> Date: Mon, 24 Jul 2006 23:08:31 +0200
> To: <grassuser at grass.itc.it>
> Subject: Re: [GRASS-user] question on georectification
> 
> I wonder why the result should be bad! If the points for rectification
> are ok, then it should fit quite well. But today I had not the time to
> finish the first map, so tomorrow evening I can report my reports.
> 
> Wolfgang
> 
> Harri Kiiskinen schrieb:
>> ma, 2006-07-24 kello 19:58 +0200, Wolfgang kirjoitti:
>>> Hi,
>>> 
>>> I've scanned some maps but as my scanner is not big enough I had to scan
>>> each quarter on its own. I've imported these scans into a xy-location.
>>> Can I use i.group to group these 4 scans together and then i.points to
>>> get directly the full map? Or do I have to project the scans
>>> individually and to combine them after rectification?
>> 
>> I did the latter and the results were not too good; even though using 10
>> to 12 control points, the overlapping parts did not sit together too
>> well. And you have to create a separate location for each scan (logical
>> though not intuitively apparent).
>> Were I to do it now, I probably would start by combining the scans
>> together with a program like Hugin (http://hugin.sourceforge.net/). It
>> has performed well in joining partial scans of large drawings for me,
>> and the tutorials do have a nice example of just this kind.
>> 
>> Greetins,
>> 
>> Harri K.
>> 
> 
>