Improvements to Live Stacking

[carlomuccini]

IT WORKED!

Last night in the middle of a 15s livestack and saving every 15 minutes it saved me!
Thank you this feature is a wonderful gift!

Carlo

[admin]

Hey, that's great!

What was it? A satellite or plane trail? Just interested to keep track as there was a suggestion a bit further up about 'auto unstack' which would be a good idea.

cheers,

Robin

[carlomuccini]

I would say a satellite because it was a single and very bright streak that crossed the image at 45° passing right above the subject in the frame. It was 6 minutes into the shot and it perfectly eliminated the annoying trail.
Excellent work!

Carlo

[timh]

Thanks once again Robin. The frame to frame drift measurement will be particularly useful for me since with a big scope and long f getting precise balance can be an issue. Tim

[admin]

Hi Tim,

this week's update takes the drift measurement a bit further and shows a plot of how the drift varies across the frame (particularly useful if you have field rotation going on)...

Screenshot 2024-07-08 133755.png (9.3 KiB) Viewed 2432 times

I am also planning to look at detecting satellite trails in the stacked image to see if I can do auto unstack, but the key there is reliably detecting the problem.

cheers,

Robin

[Gfischer]

Hi Robin,
I was wondering about the method of detecting trails or other things such as planes. Could you take a total of all pixel values in previous frame and compare to total values in current frame. If variations is above user set percentage than frame is rejected? I no nothing about programming so maybe this isn’t practical. Just an idea.

Clear skies

[timh]

Just thinking more the measurement of drift. Will it still work OK with dithering provided the dithering is only applied every third frame or less frequently? ie thinking there have to be at least two consecutive frames where the telescope is supposed to be not moving versus the sky?

[admin]

Hi,

you will definitely notice a big jump in position when a dither occurs between two frames. Since the offset is just being shown for information at the moment, that doesn't really matter. In future, if I start using that offset info for something (maybe warning that the drift or rotation is too fast for the exposure time) then I will have to keep track of when the dither happens and ignore the values calculated between two dither frames. Of course, if you are dithering after every frame then the results will be complete rubbish, only really showing you the size of the dither in camera pixels (but maybe that's interesting too!).

Going back to the idea of auto-detecting a problem and unstacking, the key really is to try to make it as automatic as possible - the more 'sensitivity' or 'threshold' parameters you have to tweak, the less usable it is. At the same time, it has to be accurate - no good if it throws out good frames or regularly fails to spot bad ones. I have some algorithms already that are pretty good at spotting lines in images, so I should be able to re-use those.

cheers,

Robin

[Jean-Francois]

Hello,

Interesting.

I have some algorithms already that are pretty good at spotting lines in images

Good ... but if you have strong diffraction from the spider on bright stars, then you will ignore all the frames.

Is it possible to ignore only a region around the lines, but not unstacking the full frame ?

Regards,
Jean-Francois

[admin]

Hi Jean-Francois,

ignoring a region around the lines is what the current 'Satellite Trail Removal' feature does, but that needs a fair bit of careful tuning of settings to get it working nicely. By looking at the final stacked image you have a better chance of detecting trails that end up being visible and ignoring those that are not causing a problem without so much tuning.

As for star spikes - I can see at least two ways around this. A minimum line length or by looking to see if any *new* lines appear in the newly processed stack.

cheers,

Robin