--verify still blind solving all-sky and ignoring CRPIX header definition

16 views
Skip to first unread message

Karsten Schindler

unread,
Jul 18, 2022, 10:00:34 PMJul 18
to astrometry
I assumed the parameter --verify essentially runs solve-field with a known RA, Dec and plate scale and see if it can improve the header when matching the field to the indices at hand. This should make it pretty quick.

In my current case, --verify attempts to verify using all indices ranging from 10% to 100% of the field size but gets zero solutions, which triggers a regular all-sky blind solve. Moreover, we are using a specific CRPIX definition not at the image center, but the result is a file with a seemingly arbitrary CRPIX location - apparently, the blind solve that is executed after --verify fails ignores any existing WCS definition.

I am having trouble to understand why our files can't be verified. The files at hand have a WCS header based on a simple TAN / gnomic projection without SIP coefficients, and the same keyword definitions as solve-field would add. The files have been verified by other means, i.e. they produce correct results in Aladin.

What am I missing, or do I misunderstand --verify completely?

Thanks -
Karsten

Dustin Lang

unread,
Jul 19, 2022, 9:44:54 AMJul 19
to Karsten Schindler, astrometry
Hi Karsten,

With "--verify", the solver detects stars in your image, and then performs the usual probabilistic test it uses to decide whether a proposed solution is good or not -- it projects reference stars into your image, and then for stars detected in your image, asks if they are surprisingly close to reference stars.  This typically only works if the WCS is very good -- like within a few pixels of being correct.  Since it's depending on matching reference stars to stars detected in your image, the --verify option has the usual Astrometry.net caveats -- the image has to contain enough detectable stars, and the bandpass must be close enough to the reference catalog that the brightness orderings are similar.

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/35c66f5f-cf1f-487d-833f-6b78c4ec0ffdn%40googlegroups.com.

Karsten Schindler

unread,
Jul 20, 2022, 2:05:47 AMJul 20
to astrometry
Hi Dustin,

Our images have enough stars and the WCS header checks out in Aladin, definitely good to a few pixels. That is why it surprises me that --verify does not work. I am using the 5200 indices. Is there any way to constrain parameters for the blind solve that commences once --verify fails?

Karsten

Dustin Lang

unread,
Jul 20, 2022, 8:24:49 AMJul 20
to Karsten Schindler, astrometry
Hi Karsten,
Interesting -- would you mind sending an example image?
For the blind solve following --verify, there is --cpulimit (CPU time to spend in seconds), but --objs (number of stars to look at) is probably better.
cheers,
dustin


Reply all
Reply to author
Forward
0 new messages