Registration fails....
 
Share:

APP 1.076 will soon be released... check the release notes !

APP 1.075 has been released !

2019 November: Complete LRGB Tutorial of NGC292, The Small Magellanic Cloud by Christian Sasse, Astronomer in Charge of iTelescope.net

2019 September: Astro Pixel Processor and iTelescope.net celebrate a new Partnership!

Registration fails... Too many "fake" stars?  

  RSS

(@schullerfred)
White Dwarf Customer
Joined: 11 months ago
Posts: 11
September 30, 2019 23:11  

Hi,
I have been using APP successfully several times with the same setup (C8 Evo and ASI 385MC) but it won't process my last capture...When I run 4) register (with pattern recognition set to "triangles") I get an error message "Registration failed on some of the frames!" - which sounds actually almost nice, since in reality it failed on ALL frames.

I'm suspecting that the problem is that it finds far too many "stars" when running 3) analyse stars: the results show typically 200+ stars, while I can only see 10 to 15 stars in each frame (these are short exposure, 10 s each). Sure, a computer can do better than my bare eye, but still, I believe that most of these "detections" are actually noise peaks. And then, trying to find pairs between two images cannot provide any meaningful solution.

I would be happy if there would be a parameter in 3) analyse stars to limit the results to a small number (say, 10 or so) of the brightest stars in the image. In principle, one needs only 3 pairs to fully describe the translation + rotation between two frames. All my light frames are towards the same field, only some small drift (because no guiding) and field rotation (because alt-az mount) can happen, no distortion and no need for mosaicing at this stage...

Any idea what parameter I can change to fix that problem?

This topic was modified 4 months ago by SchullerFred

ReplyQuote
(@schullerfred)
White Dwarf Customer
Joined: 11 months ago
Posts: 11
October 1, 2019 10:33  

Good morning...

After investigating a bit more, I understand the root cause of my problem: the true stars are too big on my images. Somehow, APP assumes that a star has a FWHM of 2 to 3 pixels, but in my case, it's more like 12 to 15. OK, the focus was maybe not perfect, but still, the size of the PSF depends on the pixel size of the camera, the telescope, and the seeing during the observations. Therefore, it would be nice to be able to keep this as a free parameter (within some reasonable limits).

Now, what can I do with my frames so that APP detects the real stars, and not the peaks in the noise?

Thanks for your help!

Fred

 


ReplyQuote
(@vincent-mod)
Quasar Admin
Joined: 3 years ago
Posts: 1108
October 1, 2019 11:04  

This sounds like APP is seeing bad pixels as stars and we have seen this before, also with someone who had out of focus data. This has been addressed and will be fixed in the next version. However FWHM of 12 to 15 is big, usually it's not that high even with a different set-up.

This post was modified 4 months ago 2 times by Vincent Groenewold - Moderator

ReplyQuote
(@schullerfred)
White Dwarf Customer
Joined: 11 months ago
Posts: 11
October 1, 2019 11:09  

That night I was using a 2x Barlow, thus working at F/20... So actually my set-up was not the same as in other nights, when APP worked well. Still, I would be happy if it could deal with "bigger" stars as well!

 


(@vincent-mod)
Quasar Admin
Joined: 3 years ago
Posts: 1108
October 1, 2019 11:28  

Yes it should be able to, so that's where the fix will come in.


ReplyQuote
(@schullerfred)
White Dwarf Customer
Joined: 11 months ago
Posts: 11
October 22, 2019 11:15  

Hello!

So, I have just downloaded version 1.075 and tried it on the same set of data. Unfortunately without success. The problem now occurs earlier, in step 3) analyse stars, where ALL the frames produced an error "analysis failed". I have tried with and without calibrating the frames (I only have darks for that night, no flat and no bias). I have also tried using "super-pixel" as debayer algorithm instead of adaptive Airy disc, so that the PSF looks "smaller" (i.e spread less pixels). None of this did fix the problem...
I would be happy to upload an example frame if you have a chance to look at it 🙂

 

This post was modified 3 months ago by SchullerFred

ReplyQuote
(@vincent-mod)
Quasar Admin
Joined: 3 years ago
Posts: 1108
October 22, 2019 11:42  

Sure, it must be a problem in the data as in 1.075 the analysis of stars was greatly improved, also to deal with hot pixels in some case. You can simply upload a stretch of one of the failed light frames here as a screenshot?


ReplyQuote
(@schullerfred)
White Dwarf Customer
Joined: 11 months ago
Posts: 11
October 22, 2019 11:48  

Sure, here's a screenshot of one frame (dark-calibrated).

Screenshot 2019 10 22 at 11.46.06


ReplyQuote
(@vincent-mod)
Quasar Admin
Joined: 3 years ago
Posts: 1108
October 22, 2019 11:51  

Excellent, thank you. So the issue in this light seems to be a very high noise level and low signal. There also aren't many stars visible, these things combined make it almost impossible to properly register.


ReplyQuote
(@vincent-mod)
Quasar Admin
Joined: 3 years ago
Posts: 1108
October 22, 2019 11:56  

Could it be that you have to optimize your gain settings etc still? It must be possible to get a much higher signal over noise with this camera.


ReplyQuote
(@schullerfred)
White Dwarf Customer
Joined: 11 months ago
Posts: 11
October 22, 2019 11:56  

Here is the same frame but without calibration, and with a slightly different stretch. I agree that there are not many stars, but I still expect the software to detect at least 10 or so...

Screenshot 2019 10 22 at 11.54.34


ReplyQuote
(@vincent-mod)
Quasar Admin
Joined: 3 years ago
Posts: 1108
October 22, 2019 12:00  

Yes that seems better, although the noise is still too high for what I would expect. Could you upload some frames to our servermaybe? It is challenging data and that is great for Mabula to see if he can actually make that work. Login/password: appuser and please create a directory with your name on it. Much appreciated!


ReplyQuote
(@schullerfred)
White Dwarf Customer
Joined: 11 months ago
Posts: 11
October 22, 2019 13:52  

OK, I uploaded one dark frame and 4 lights in directory SchullerFred.

As I mentioned earlier in this thread, the stars have a PSF as big as 12-15 pixels, due to using a 2xBarlow and certainly not perfect focus. I think that the star detection step is one of the most critical, and it would be nice to have access to a few input parameters, such as the expected typical size of the stars, to give a hint to the programme...

 


ReplyQuote
(@vincent-mod)
Quasar Admin
Joined: 3 years ago
Posts: 1108
October 22, 2019 19:30  

Excellent, thanks for that and Mabula will have a look.


ReplyQuote
(@schullerfred)
White Dwarf Customer
Joined: 11 months ago
Posts: 11
November 30, 2019 23:48  

Hello,

did Mabula have chance to look at some of my problematic frames?
Last night I had clear weather but terrible seeing. I tried to get some images (without Barlow) but again I have the problem that Analyse stars fails on *all* the frames. Here's an example screen shot. The stars look big (because of the seeing) but there are quite a few in the image that APP should be able to detect...

Is there any hope to find a way to process these images?

Screenshot  2019 11 29


ReplyQuote
(@vincent-mod)
Quasar Admin
Joined: 3 years ago
Posts: 1108
December 3, 2019 13:05  

Sorry, Mabula has been abroad and unfortunately I didn't receive word yet for when he can attend the forum. Should be soon, I messaged him again. @mabula-admin

This post was modified 2 months ago by Vincent Groenewold - Moderator

ReplyQuote
Share: