Using the Astrometry.net's output as a 'first guess' for more accurate image solving tools such as WCSTools

[Kshitij Duraphe]

The 2010 publication detailing Astrometry.net says: There are several automated calibration systems that refine the astrometric calibration of an image to produce a high-precision alignment to a reference catalog given a good first guess

What tool would be the 'best' for the job, assuming that this is exactly what I want to do? A thread from 2013 (https://groups.google.com/g/astrometry/c/Ky1er5VpY3c) in this very group used WCSTools, but I'm interested to know if the community has developed anything better. To be very specific, I am using images of the night sky taken from Earth with cameras ranging from cell phones to high-quality cameras, but all the images are wide-field images.

Also, do there exist any tutorials for these kinds of tools? I'm not sure how to use the output of Astrometry.net as the input to WCSTools...

[Dustin Lang]

Hi,

It's maybe worth saying what Astrometry.net produces and what is missing.  Astrometry.net starts by matching 3- or 4-star features in your image to a reference catalog of known stars.  It then uses 10+ more stars to confirm that match.  So normally we have 10-20 stars matched, sometimes many more.  Starting with that, we can also compute polynomial distortion terms to give a more precise match between image stars and known stars.

This is all done image-by-image, and always matching images to reference catalogs.  For some images (but not wide-field ones), they can be deeper than the reference catalogs, so aligning image-to-image could improve things.  And doing simultaneous fitting (including assuming that some parts of the distortion are constant -- like Scamp done) could also improve overall results for a large image collection.

For wide-field images, I would say the bigger issue would be getting good distortions, since they tend to be larger for wide-field (especially consumer-grade) images.

cheers,

dustin

--
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 on the web visit https://groups.google.com/d/msgid/astrometry/60226fc8-58a2-474c-b808-b5106993b6cbn%40googlegroups.com.

[Kshitij Duraphe]

Thanks for the quick reply! Just to clarify, by 'simultaneous fitting', you mean fitting an image with astrometry.net and then fitting it with WCSTools, right? Also, what is the default polynomial order used for the SIP fitting? Is this something I have to manually tweak? It looks like tweak-order is something I can set to enable polynomial fitting. What could good starting values be? I'm personally thinking about something that maximizes the average correlation in the .corr file, but I may have misunderstood the contents of the correlation file.

Also, I'm not sure what Scamp is - is it the name of the author of a publication?

[Dustin Lang]

Hi,

I've never used WCSTools, so I can't speak to that.  Scamp is https://www.astromatic.net/software/scamp/

I believe the default SIP polynomial order is 2.  If you have enough matched stars, you could go up to 3, 4, 5.  Beyond that, I'd be wary.

cheers,

dustin

To view this discussion on the web visit https://groups.google.com/d/msgid/astrometry/1a1328a2-1dd8-4132-9328-d3434b27c3cen%40googlegroups.com.

[Kshitij Duraphe]

Thanks Dustin. Please excuse my naive questions - I'm still very much a beginner. My overall goal is to estimate the scale of objects in the night sky given some assumptions about their altitude, from consumer camera images. My approach is to 1. platesolve the background starfield to determine orientation and 2. 'un-distort' the image. From what I understand, the 'un-distort' procedure should involve finding out the coefficients of the SIP or the PV polynomial (which is why I was looking into WCSTools) and then finding out its inverse polynomial, so I can apply this undistortion to the image. I believe this then maps the image onto the celestial sphere. (i.e. converts the arbitrary distortion to a radial distortion). Is my understanding correct?

If it is, then I can use my undistorted image and estimate the scales of objects in it, now that the distortion is something I know. I was considering using the corr.fits file generated by astrometry.net to evaluate my undistortion, but I'm not sure if this is something I should be doing.

[Dustin Lang]

What do you mean by

"My overall goal is to estimate the scale of objects in the night sky given some assumptions about their altitude"

... are you talking about objects within the atmosphere?

If you just want to know the angular scale of the image pixels (eg 1 pixel = 0.0437 degrees), you really don't need to do all this detailed distortion measurement, the initial results from Astrometry.net will be good enough.

One other thing to note: if you want the *azimuth* and *altitude* that the camera was pointing at (how far above the horizon, in which direction), then you also need to know the camera's location and time when the image was taken, the star field alone will not tell you that.

cheers,

dustin

To view this discussion on the web visit https://groups.google.com/d/msgid/astrometry/4d6140e1-0b8a-4749-8254-4f81bd451768n%40googlegroups.com.

[Kshitij Duraphe]

Yes, I'm talking about objects in the atmosphere. Specifically, I'm trying to estimate the size of the STEVE phenomenon from images posted on the internet. I platesolve STEVE images directly since astrometry.net is usually good enough to get a starfield it, but I don't account for the overall distortion. I know the approximate location of the camera (to within ~1 mile, which is fine, since STEVE tends to range for thousands of miles) and the time of the image.

I don't quite understand if 'undistorting' the images solved by astrometry.net is the right approach to provide an image mapping I can calculate structure scale in, so I wanted to ask.

[Dustin Lang]

Hi,

The World Coordinate System results that Astrometry.net produces is a map between pixels and RA,Dec positions on the sky.  One property of that mapping is the degrees-per-pixel.  I think that's all you need for your project, isn't it?  Or do you also need to know the altitude of the camera pointing (angle above the horizon)?

The WCS can either be a simple tangent-plane projection, or it can have a correction for image distortion using SIP polynomials.

cheers,

dustin

To view this discussion on the web visit https://groups.google.com/d/msgid/astrometry/3b8bd73f-19ea-4623-8cb9-c58dc4a900d0n%40googlegroups.com.

[Kshitij Duraphe]

Ah, I see. I thought that the WCS mapping was without the image distortion correction (the tangent-plane mapping) and the SIP polynomial coefficients were stored in the header, and the user had to manually undistort the image using these coefficients. In that case, I feel comfortable going ahead with the Astrometry.net solution without further processing. I believe I only need the degrees per pixel. I know the altitude of the camera, and where its pointing at.

Thanks so much for your help!