Solving wide fields with many stars (solution sometimes latches onto a small part of image)
45 views
Skip to first unread message
Serge
May 18, 2018, 3:21:49 AM5/18/18
to astrometry
Any recommendations to improve the robustness of solutions for wide-fields (~20 deg) with many sources (800+).
A the moment there are occasions where the solver matches 50/140 stars to 800+ sources but all the matches are bunched to one side of the image and the opposite side is very poor.
While some solutions work very will with 27/28 stars matched because matches are spread out evenly.
One solution I see is to extract the sources manually and provide only say 30-40, which are well distributed over the field.
However if source position is very uncertain then this thoughts out a lot of useful information. Ideal would be to go the other way and ask astronomer to increasing star counts (or let user set it) and increase tolerances if needed until it can match some user specified percentage of the sources (say 80%).
Going that way would also allow much better estimation of the true fit accuracy. Because it is possible to match say 20 stars with 800 noisy sources with low residuals by simply selecting the right sources, but the residuals do not indicate how good the solution is. Using most of the stars should improve the fit and allow better estimation of accuracy.
This would also apply to narrow fields that have a lot of sources.
Perhaps this already possible using the current offline solver?
Best Regards,
A the moment there are occasions where the solver matches 50/140 stars to 800+ sources but all the matches are bunched to one side of the image and the opposite side is very poor.
While some solutions work very will with 27/28 stars matched because matches are spread out evenly.
One solution I see is to extract the sources manually and provide only say 30-40, which are well distributed over the field.
However if source position is very uncertain then this thoughts out a lot of useful information. Ideal would be to go the other way and ask astronomer to increasing star counts (or let user set it) and increase tolerances if needed until it can match some user specified percentage of the sources (say 80%).
Going that way would also allow much better estimation of the true fit accuracy. Because it is possible to match say 20 stars with 800 noisy sources with low residuals by simply selecting the right sources, but the residuals do not indicate how good the solution is. Using most of the stars should improve the fit and allow better estimation of accuracy.
This would also apply to narrow fields that have a lot of sources.
Perhaps this already possible using the current offline solver?
Best Regards,
Eric Sibert
May 18, 2018, 4:33:01 AM5/18/18
to astro...@googlegroups.com
> Any recommendations to improve the robustness of solutions for wide-fields
> (~20 deg) with many sources (800+).
I'm often doing local solving on wide field (20°x15°) picture. As I'm
> (~20 deg) with many sources (800+).
doing it locally, I first adjust the star level detection so to have
between 100 and 200 stars. Second, after a first solving, I perform it
again using the refine option. (Third, I built my own indexes to
better match my sensor -DSLR- sensitivity :
https://groups.google.com/d/msg/astrometry/hGa74stffzE/owshZMxJAgAJ).
Se also:
https://groups.google.com/forum/#!searchin/astrometry/sibert%7Csort:date/astrometry/IUZYaJiLWwQ/Tlh_zuYcAAAJ
Eric
Reply all
Reply to author
Forward
0 new messages