Re: Updating RA/DEC fits headers with true location?

[Stupendous Man <richmond@stupendous.cis.rit.edu>]

  Robert asked:

> So, I have a bunch of stars with known astrometric positions 
> (RA/DEC) and known frame positions (x,y).  I'm sure there is a 
> 'proper' way to calculate the center of the image, but I'm not 
> much on spherical geometry yet.

  Arne answered:

> If you have the transformations to calculate RA,DEC from a given x,y
> then you just use those equations with the center pixel as input.
> It sounds like you are one step earlier, though; having matched
> a set of stars with known RA,DEC and with CCD x,y how do you
> determine the transformation?  I'd suggest looking at the
> astrometry chapter in Berry/Burnell, The Handbook of Astrnomical
> Image Processing, Willmann-Bell.   Good book to own in any case.
> There is a nice set of routines in slalib to do the math.

  If Robert is indeed at the state Arne mentions -- having two
matched sets of stars with both (x,y) and (RA,Dec) -- then 
I second Arne's recommendations.  But let me add a few more
options:

   - as Robert suggested, a very quick but probably fine method
         is to pick two stars close to the center of the (x,y)
         system and do a simple linear interpolation.  That is,
         if the central x value lies 30 percent of the way from 
         x1 to x2, and RA corresponds nearly exactly to x, then
         the central RA position will be about 30 percent of the
         way from RA1 to RA2.

   - you might also look at the "match" package
     
            http://spiff.rit.edu/tass//match/match.html

         which will find the matching items on the two lists
         (which Robert may already have done) _and_ provide
         a mathematical description of the transformation
         (which Robert would like).  You can choose one of 
         three different transformation models, depending on
         your field of view and need for correcting large-scale 
         aberrations.

  It's been my experience that the Mark IV images, over 4 degrees on
a side, will drive you crazy no matter what :-/  What I mean is,
even if you find the central (RA, Dec) position of the image
perfectly, residual differences between the mathematical model
given in the FITS header and the actual positions of stars on the
image will lead to large offsets between the position predicted
by a simple image viewing program like DS9 or fv and the actual
stellar position in the image.  I'd love to be proved wrong, so please
do let me (and the entire mailing list) know if you have found a 
good solution to this problem.

                                                Michael Richmond