Source detection

[Gaetano Zarcone]

Hello everybody

I'm trying to find the ra,dec of the stars in my image and I'm trying to compare these with the ones obtained from the catalog reading the corr file.

The goal is to validate my source detection algorithm.

To achieve this, I perform two procedure:

1 - compare the index_ra, index_dec with the field_ra, field_dec and what I find is

2 - with my source detection algorithm I find the pixel coordinates of the stars, then I convert them in ra,dec using wcs. Finally I compare these with the ra,dec retrieved from catalog and what I obtain is

You can see that there is a problem in the detection algorithm, just seeing the axis scale (hundreds of arcseconds vs tenths of arcseconds)

My question is: which procedure does Astrometry perform to detect sources in the image?

Thank you for your time

[Dustin Lang]

Hi,

Things to check:

-Are you using the SIP coefficients in the WCS header correctly when doing the X,Y to RA,Dec conversion?

-How are you deciding which stars match?

The Astrometry.net source detector is in image2xy.c and the core algorithm is in simlpexy.c , which has a description of the algorithm

https://github.com/dstndstn/astrometry.net/blob/main/util/simplexy.c#L22

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/CAGm%2BbRWNdaUWxiWvafy%2BrU3qdsavtr%2BpW%2BBw0oddt8P_kkKF%3Dw%40mail.gmail.com.

[Gaetano Zarcone]

Hi Dustin,

thanks for the reply

1- Yes, I solve the image with the --crpix-center options and with a tweak order = 3. In this way, when I convert the x,y to ra,dec the wcs has already the SIP coefficient and the distortion is considered

2- I use the stars with an index_match > 0.99 (There are more than 1500 stars in the field of view, so I can "push" the index_match threshold to 0.99)

Now I take a look at the code

cheers,

Gaetano

[Gaetano Zarcone]

Other clarification

the stars in my images are not "dot" but are stripes given the 15 seconds of the exposure time and the fixed field of view.

Hope beeing clear

Cheers,

Gaetano

[Dustin Lang]

Ahh, yes, in that case, you might find that simplexy (or your algorithm!) identifies one of the endpoints, or the middle, as the source.  If you're running solve-field with plots, check the -objs.png plots.

[Gaetano Zarcone]

Ok

thanks a lot Dustin

[Gaetano Zarcone]

Hi Dustin,

I have another question and I hope for help.

When I convert the x,y coordinates in ra,dec I use the Astropy allpix2world routine as follow:

               

                  ra,dec = wcs.all_pix2world(x,y,origin)

The last input, according to the Astropy definition, is "the origin, that is the coordinate in the upper left corner of the image. In FITS and Fortran standards, this is 1. In Numpy and C standards this is 0".

I don't understand this last definition, since I use the FITS image to retrieve my coordinates but the algorithm is in a python environment (so Numpy I guess).

Which one should I choose? What do standards mean?

Thanks for your help

Cheers

Gaetano

[Dustin Lang]

Hi,

If these are pixel coordinates that you have produced for your sources, then the question is, if there is a star centered in the first pixel in the image, what are its coordinates -- (0,0) ?  (1,1) ?  The FITS convention is that it is (1,1), so you would use origin=1.

cheers,

dustin

[Gaetano Zarcone]

Hi dustin

Thanks for reply

Now it is clear

Cheers

Gaetano