astrometry from a serie rather than single shot

[Jeremie Vaubaillon]

Hello,

Here is a probably strange question: instead of feeding astrometry.net with a list of star sources taken during a single shot, would it be possible to feed it with a time serie of (x,y) coordinates of a limited number of stars (e.g. <10) as they drift in the FOV, and the corresponding (ra,dec) for each of them (identified e.g. manually)? If not, is there any software that would do such a trick please?

thanks :-)

Jeremie

[David W Hogg]

This is a great request, and should in principle be "more robust" than Astrometry.net as it is. It would also be useful for some of our NASA and ESA customers. We don't currently have a solution for this.

David W. Hogg | http://cosmo.nyu.edu/hogg/

--
You received this message because you are subscribed to the Google Groups "astrometry" group.
To unsubscribe from this group and stop receiving emails from it, send an email to astrometry+...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/astrometry/ed256233-7229-4b5a-8aa4-c869997700e5n%40googlegroups.com.

[Dustin Lang]

Hi,

You want this for getting a detailed distortion model?

The "fit-wcs" program in the Astrometry.net package takes matched X,Y and RA,DEC files and fits a SIP WCS.  If you had a sequence of images, you could just append all the X,Y,RA,DEC values.

However, to do this properly, I think you would want to re-fit the individual-frame WCS solutions.   We don't have that.  With ~10 stars it's not super well constrained.

cheers,

dustin

--

[Michael Roberts]

I’d be keen to look into this for SkySolve (https://github.com/observerly/skysolve).

I have a lot of questions though. What is the use case and why is this needed?

On 16 Jul 2025, at 07:09, David W Hogg <dwh...@gmail.com> wrote:



To view this discussion visit https://groups.google.com/d/msgid/astrometry/CAEZJrfpBVSxzNYvNyviRhOBjCJRorYFQjpS%3DsgFswOFLTpoSog%40mail.gmail.com.

[Jeremie Vaubaillon]

ok thanks for your replies. Too bad it doesn't exist yet...

For the readers who didn't really get the point: if you have fixed camera with wide FOV you won't get many stars in the FOV, so to find a (x,y)->(ra,dec) transformation will be impossible (or at least not well constrained so the solution will be crap...). The idea is to map the (alt,az) space, knowing the (ra,dec) of a star, and following its (x,y) as a function of time. You'll derive a (x,y)->(alt,az) transformation, which in turn can be converted into (x,y)->(ra,dec) for a given time. In the end you will get a decent astrometry solution even with 10 stars in the field of view.

@David Hogg: that would be a nice student project, especially if there is a need from the space agencies. Please contact me off line (vaubaill [at] imcce.fr) if you're interested for additional information.

Jeremie

[Jeremie Vaubaillon]

Hello,

not sure if my previous reply went through... so here it is again...

@Dustin: yes I'd like a distorsion model, but also the ra,dec of detected objects.

@David: maybe that'd be a good student project (?) Plz contact me offline for further discussions (vaubaill [at] imcce [dot] fr)

For those wondering the use: suppose you have a wide FOV fixed camera with only a few stars visible (e.g. for meteor observations) and only a handful of stars detected, then it will be nealry impossible to get a correct astrometric solution. So the meteors can't be triangulated. Now, considering a known and detected star, you can follow it during the night, so you have a serie of (x,y)=f(t) and corresponding unique (ra,dec). The idea is to work in the (alt,az) frame: given a single (ra,dec), you can work out the (alt,az)=f(t) and find a transformation between (x,y) and (alt,az). Once this is done, for a given time, you can know the (ra,dec) of any (x,y).

Well, I guess it would do the trick if it would be possible to change the frame in which astrometry.net is working, but of course I have no idea how complicated of a work this represents..

Jeremie

On Wednesday, July 16, 2025 at 12:57:14 PM UTC+2 dstn...@gmail.com wrote:

[Jeremie Vaubaillon]

Hello,

my previous replies didn't get through, so here it is again.

The use case is the following: I have a wide FOV camera for meteor monitoring. The number of detected stars is very low (<10) so astrometry.net will not converge and/or provide a solution that makes sense with such a small number of stars. The idea is the following: map the (alt,az) space thanks to a known stars for the the (ra,dec) are known and can easily be converted into (alt,az) for a given time. This way you have a serie of (x,y)=f(t) that you can map to a serie of (alt,az)=f(t). Once you have such a transformation for all times, you can go back to (ra,dec) for one specific time.

Of course I have no idea how complex such a problem would be to solve, but I'll have a look at SkySolve.

@David: that would be a great topic for a student :-)

Jeremie

[Dustin Lang]

Hi,

Okay, so you have a set of stars with known RA,Dec, and a set of known times t_i and pixel positions x_i, y_i.  From RA,Dec, t_i and your GPS location, you can compute alt_i, az_i.

Now you have x_i, y_i, alt_i, az_i.  Choose some arbitrary time t_r.  Take t_r, your GPS location, alt_i, and az_i, and compute ra_i, dec_i.

Now you have x_i, y_i, ra_i, dec_i -- a bunch of predicted RA,Dec positions that would project to the x_i,y_i locations if you observed at time t_r.

Feed that list of x_i, y_i, ra_i, dec_i into the Astrometry.net fit-wcs routine, or your favourite other fitting routine -- it's just a non-linear least squares fit, no decisions to make.

Or do I have something wrong?

cheers,

dustin

To view this discussion visit https://groups.google.com/d/msgid/astrometry/490a7241-041c-4b58-85f0-c72226baa4d8n%40googlegroups.com.

[Jeremie Vaubaillon]

Sounds like it! Thanks++ I'll give it a try :-)

Jeremie