About scaling, transformation and rotation of FITS file
jeffrey wrote on Feb 02, 2009
I would like to ask a questions about scaling, transformation and rotation of FITS file, actually I got ~10 FITS which has a small shifting, rotation and scaling with each other, and in those images, only one object has moved while other background stars should remain rest. Now I would like to find the amount of movement of that object.
I know I should find several reference stars, rotate, rescale the FITS, but I dont know which tasks I could use.
I have spotted the geomap, geoxytran but can the tasks also do the scaling?
cause I am not sure that answer will really be the one I need.
Also do I need to generate a WCS coordinate? Will it be better?
Thank you very much
Sorry for asking dumb questions again.
I know I should find several reference stars, rotate, rescale the FITS, but I dont know which tasks I could use.
I have spotted the geomap, geoxytran but can the tasks also do the scaling?
cause I am not sure that answer will really be the one I need.
Also do I need to generate a WCS coordinate? Will it be better?
Thank you very much
Sorry for asking dumb questions again.
Francisco Valdes wrote on Feb 02, 2009
This is not a dumb question. There are different levels of analysis and difficulty. At its most detailed this is the kind of work astrometrists do with extreme care. I am not an astrometrist so I will just give some general thoughts.
The first step is to measure (centroid) a number of stars near your target; that is the ones you believe are fixed and form your reference system. There are various tools that can find the centroids. Interactively you could use imexamine and the 'a' or 'r' key. Or you can make a list of the x/y values using a cursor and then feed that into imcentroid.
Once you have the positions of the stars you would average them (possibly with weights) to get a fiducial coordinate. Finally you would centroid the target star and compute the distance from the fiducial coordinate. The sets of distances from the fiducial coordinate in each image provides your source motion.
Note that I have described this in terms of using the image x/y coordinates. That is fine if the reference stars are relatively close or the images are small with minimal distortions. The same analysis can be done in celestial coordinates. Then you would have to calibrate the world coordinates of each image using various tools including ccmap to fit an astrometric solution. You can then use wcsctran to convert your x/y values to ra/dec. Then you do essentially the same thing to find distances in terms of arcseconds (remember to use spherical distances and not just delta ra and delta dec).
I hope this gives you some guidelines and useful suggestions.
Frank Valdes
The first step is to measure (centroid) a number of stars near your target; that is the ones you believe are fixed and form your reference system. There are various tools that can find the centroids. Interactively you could use imexamine and the 'a' or 'r' key. Or you can make a list of the x/y values using a cursor and then feed that into imcentroid.
Once you have the positions of the stars you would average them (possibly with weights) to get a fiducial coordinate. Finally you would centroid the target star and compute the distance from the fiducial coordinate. The sets of distances from the fiducial coordinate in each image provides your source motion.
Note that I have described this in terms of using the image x/y coordinates. That is fine if the reference stars are relatively close or the images are small with minimal distortions. The same analysis can be done in celestial coordinates. Then you would have to calibrate the world coordinates of each image using various tools including ccmap to fit an astrometric solution. You can then use wcsctran to convert your x/y values to ra/dec. Then you do essentially the same thing to find distances in terms of arcseconds (remember to use spherical distances and not just delta ra and delta dec).
I hope this gives you some guidelines and useful suggestions.
Frank Valdes
jeffrey wrote on Feb 02, 2009
Really thanks for your detail explanation, it helps me know more about what I am doing, luckily, it seems that I am on the right way.
But I got troubles in getting the errors from the results again...
Seems the tasks do not give results of error, but I really what to know how accurate the result is.
Do the error just similar to the one in imcentroid so that I can calculate it without scripting the whole task?
But I got troubles in getting the errors from the results again...
Seems the tasks do not give results of error, but I really what to know how accurate the result is.
Do the error just similar to the one in imcentroid so that I can calculate it without scripting the whole task?
Francisco Valdes wrote on Feb 02, 2009
I'm afraid you will have to do the error analysis yourself based on the centroid errors of the individual sources. From that you can get a formal error in the average reference system and, in combination with the error in the centroid of the target source get an final uncertainty.
Best wishes,
Frank Valdes
Best wishes,
Frank Valdes
Last post on Feb 02, 2009