Too few stars detected in Star Analysis

An addendum from my side: I made another integration of the red channel as a test and now the star analysis finds >2000 stars. Meanwhile I could combine a 2-panel mosaic of the red chanel. I'll do the same for the other channels and hope it will work. I'll report it here asap...

@pete_xl Hi Peter,

In tab 3 you can increase the number of stars used to 2000 or more. Then in tab 4 you can set the scale start at 5 and scale end at 10 or more and registration should work.

HTH, Wouter

Hi Wouter and thanks for the support! 

I already had set stars to 4000 and scale to 5 and 15 (I know APP quite well for a long time and have made some mosaics already).

I have to report that the problem continues. In the first star analysis of red APP had counted +2,000 stars and the registration went fine. But it does not work with the green channel. The image has as many stars as red but APP only counts 51! What is even more strange is that I loaded red once more and APP counts only 475 stars now instead of 2,500 before in the same image!

This is the red chanel that had >2,000 stars before:

It has only 475 in the actual analysis.

And this is green that has only 51 stars (according to APP 🤨 ):

I cannot see any significant difference between the two that could lead to such different results.

And this is the green channel from the second panel that counts 1.599 stars:

At least visually has less stars than the other two.

This is really weird and I'd really liked to solve this because it would be a really great mosaic. The luminance channel looks overwhelming. Here it all was flawless:

Cheers Pete

Hi Wouter, it's me once more 😀 .

I installed an old version (1.076beta) I found in my download folder. It delivered nearly 4,000 stars instead of 51 from the same file!!

The registration works fine with an rms of 0.1!!! But I get this error message when I save the registrated files.

Nevertheless the other steps including the image integration seemed to work fine at the first look. But the image integration result is a piece of star art:

I will now reinstall the actual beta and try to integrate the registered files there

My conclusion is that there is some kind of star analysis problem in the actual beta.

Cheers Pete

...here we are! The integration came out nice in 1.083 stable_

😀

@pete_xl Beautiful dusty mosaic!

Thank you Wouter 😀 ! But this is only the raw stretched data. They look very promising though.

But I want to inform you that there really seems to be a bug in V2.0.0 beta4. In V1.76 and 1.83.4 the star analysis in the already integrated single master panels is flawless and finds thousands of stars. In 2.0.0 beta 4 it delivers only a few (10-50). As an example her the results of the analysis of the same green panel that came out of  V1.83.4 and 2.0.0 beta 4:

V1.83.4:

V2.0.0 beta4:

All the best

Pete

Dear Peter,

Thank you very much for informing us of the problem with the beta on your data.

We have adjusted star analysis a bit in the beta, since we are always trying to improve it and make it more compatible with more data. The changes in the 2.0.0-beta do actually find more stars and more reliably in most datasets in our experience and we have a lot of test data. To make it work well again on your data, we really need to see the data ourself.

Can you upload some individual lights that cause the problems, so we can test and improve things?

https://upload.astropixelprocessor.com/

username: upload

password: upload

Please make a folder with your name and let me know once uploaded 😉

Mabula

Dear Mabula,

I am very sorry but I missed all messages here and in the parallel thread Too few stars detected in Star Analysis due to a problem with my email software.

I understand that you have solved the issue in the new Beta meanwhile and so you probably do not need my data anymore. Otherwise please give me a short notice.

Peter

P.S I love your Software!

https://www.astrobin.com/users/pete_xl/

Hello,

I am wondering whether the star detection issue has been addressed.

Recently I am working on a 2x6 panel mosaic project of the Horsehead-M42 region. I am facing the same issue. To be more specific, the screenshot shows the integrated Luminosity data which, according to APP 2.0 beta 6, has 161 stars. 

@jasonchang Hi Chang,

I'm afraid I was in a hurry and had pasted the wrong link in my answer above. There was another guy a while ago in another thread who had similar problems and if I remember it right @mabula-admin answered that he had identified the problem and will release a new version soon to correct the bug. I do not remember which thread it was.

By coincidence I had the same problem today again with  APP 2.0 beta 6 when I registereds several master stacks. They all have thousands of stars but registration failed due to too few stars in the star count. For example the SIII file shown below has 2821 stars when stacked alone. In a group together with other files for registration the star analysis procedure only counts 60. I have temporaily installed APP 1.083.4 which does the job without problems.

@mabula-admin

Hi Mabula, I will upload some of the files of the list below as soon as possible...

CS Pete

@pete_xl Thanks for the followup Peter. Unfortunately 1.083.4 does not run on M1 Mac, so it looks like I'll have to wait until the next release.

@mabula-admin Dear Mabula,

I have just uploaded 3 of the files that fail in star analysis (only 60-250 stars)and therefore also in registration. The folder name is pete_xl.

Peter

Posted by: @pete_xl

@mabula-admin Dear Mabula,

I have just uploaded 3 of the files that fail in star analysis (only 60-250 stars)and therefore also in registration. The folder name is pete_xl.

Peter

Thank you Peter @pete_xl,

I will test and improve things 😉 using your data. I think it should be improved in beta7

Mabula

Posted by: @jasonchang

@pete_xl Thanks for the followup Peter. Unfortunately 1.083.4 does not run on M1 Mac, so it looks like I'll have to wait until the next release.

Hi @jasonchang,

Can you upload some data as well so I can test and improve the star analysis/detection code ?

Please upload it here:

https://upload.astropixelprocessor.com/

username: upload

password: upload

Please make a folder with your name and let me know once uploaded 😉

Mabula

@mabula-admin 

That would be great, Mabula.

I just encountered the same problem with foreign data stacks (excellent CCD data from 2007). So it should not be an inherent problem of my data.

Peter

Hi @mabula-admin

I uploaded two stacked subs. L-?-2_Luminance-session_1.fits are stacks coming straight from AAP. L-?-2.fits are slightly cropped in PixInsight. The folder name is JasonChang-Too-Few-Stars

Thanks

Hi @mabula-admin,

its me again with an update on the star analysis problem. In Version 2.0.0 beta8 amd64 (and new computer) with my actual data I again observed that the number of stars found is too low.  

In the star analysis of star rich master integration files of NGC 7380 the star analysis (with "automatic #stars target" checked) delivers very low star count results for input values below 2,000 stars (only for the Ha image!). Registration fails here.

Up to an input value of 8.000 stars the results are higher, registration works but the numbers are still too low.

From 8.000 stars up to 10.000 stars the results are consistent and similar to the results of the SII star count.

With "automatic #stars target" unchecked APP delivers this result:

CS Pete

Hi @pete_xl,

Thank you very much, this issue is now the first on the list to be fixed for the next release. I do know why this issue happens already and i will try to make the star analysis more robust here 😉

Please accept my apologies for not having this fixed earlier, but we had to prioritize on making 2.0.0 work reliably on both macOS and Windows first.

Mabula

Hi @mabula-admin ,

thank you for your response! I'm glad you found the cause for the strange behaviour. I'm looking foreward to the new 2.0.0 stable and wish you a great start into 2022!

Pete

I am also having this issue using Beta 9, I can't attach data but I am very under sampled at 4.76*/pixel and it only seems to be an issues using the HA filter.

Hi all,

I am working on this now and I have it almost fixed, we will probably release beta10 with this fix today or tomorrow 😉

Mabula