[Solved] APP 2.0 beta 41 Registration error?

so, turning them all in LR and exporting as tiff, loading in APP works. Please comment if this will stay like that, or if APP gets improvements to get it right? Even when switching the auromatic rotation off in camera stuff like this will occur I guess.

Hi Stefan @elgol,

Thank you very much for sharing the issue.

How many stars did you detect with 3) Analyse stars ? And why are you using Triangles on such data? If you can detect 500-1000 stars, you must defnitely set the descriptors to quadrilaterals and even better to automatic. The issue message indicates to me that the field of view becomes infinite after registering somehow... It could be that you have many stars and triangles is too blame, or you have little stars detected and dynamic distortion correction becomes unstable.

As noted in the release notes, 2.0.0-beta39 includes a whole new registration engine. I have tested the new engine on a very wide variety of data and all is smooth on those datasets. I can register many more big mosaics with the new engine successfully. Of course, a new engine, unfortunately can have some new little bugs which are only triggered in certain cases. Clearly this is such a situation that I have not seen yet among our data that we use for testing and we really want to solve this of course 😉

Can you share the data on which this problem happens so I can test this in the coming days?

https://upload.astropixelprocessor.com/

username: uploadData

password: uploadTestData

Please make a folder with your name and issue like: Stefan-register-issue

and upload the data there and let me know once done. I will fix it as soon as possible.

Mabula

@mabula-admin Hi Mabula, am uploading a third of the data. new home, bad internet for now. But 4 each rotation shoulf suffice though.

I tried of course settings. Had quadrilaterals and auto used before. just now 10.000 stars at auto, same problem. Before I had 1600.

I do 2 stacks now, one before one after the flip. THE SAME!! Astonishing. The stars in the region around rho oph look like doubles:

rms error here as bad as before.

That was the only good night at our place in La Palma, so I was awake until morning. Hope we can solve this. The Sigma 200mm f/2.0 seems to optically very good.

rms reg error is always bad as below 0.3. I switched of distortion correction, error jumps to 0.75. using mosaic, I guess I tried what I could think of. stars stay bad around rho oph.

data is uploaded

Hi Stefan @elgol,

Thank you very much for the upload, i see the issue indeed. I also see that beta38 can do it properly. I will investigate thoroughly why it is failing. I already suspect the cause, i think one of the optimizations i made in beta40 is to blame. If so,  I will roll it back. I will get back to you as soon as I know more 😉

Update, beta39 also does not work properly, so it is definitely something in the new registration engine that is not sitting well with data like this where the star population is so crowded.

I will dive into the code to fix it 😉

Mabula

@mabula-admin Glad to read this Mabula. Was really strange at first... Hey it's gonna be 9 years using APP soon...

Great Stefan, Yes indeed, in 1 year time there must be a big party !!!

Please use beta38 for now, that seems to do the registration fine. I will release beta42 in the coming days with several other fixes. It will not yet include a fix for this problem, because I have not located the problem yet. But rest assured, this fix is at the very top of my list. This should work properly.

Mabula

Dear Stefan @elgol,

I have released 2.0.0-beta45 today, which has fixed the registration issue with your dataset.

It was indeed caused by a change in the new registration engine and your dataset, because it has quite a precise pierflip where the Field Of View is nearly identical except for 180 degree rotation, triggered an exception where APP turned off dynamic distortion correction by mistake. 

I have fixed it and the registration works flawless again now on this dataset (and also still on other test datasets) 😉

You can download 2.0.0-beta45 from the downloads section.

Mabula

Posted by: @mabula-admin

↑

Dear Stefan @elgol,

I have released 2.0.0-beta45 today, which has fixed the registration issue with your dataset.

It was indeed caused by a change in the new registration engine and your dataset, because it has quite a precise pierflip where the Field Of View is nearly identical except for 180 degree rotation, triggered an exception where APP turned off dynamic distortion correction by mistake. 

I have fixed it and the registration works flawless again now on this dataset (and also still on other test datasets) 😉

You can download 2.0.0-beta45 from the downloads section.

Mabula

You are most welcome @elgol,

Is all working well on all the data now? Were you able to test it yourself in beta45?

Thanks,

Mabula

Posted by: @mabula-admin

↑

You are most welcome @elgol,

Is all working well on all the data now? Were you able to test it yourself in beta45?

Thanks,

Mabula

Yes, found some time now to reprocess with 2.045. Cannot find a fault!

Thanks again Mabula

Great Stefan @elgol,

Thanks for the feedback !

Mabula