New Feature : Moon Mosaic Planner

[Borodog]

Robin,

Would it be possible to add a checkbox to the mosaic planner to automatic refocus between panels using the last successful autofocus settings?

Thanks,

Mike

[admin]

Hi Mike,

it's certainly possible - the whole lot runs via the sequencer anyway, so I can arrange to have any reasonable set of sequencer steps in there. I'm surprised that the focus would drift out that fast though!

cheers,

Robin

[Borodog]

It's not that fast. It can take of order an hour for the whole mosaic to execute.

Mike

[admin]

Hi Mike,

ok, lots of panels I guess in that case and a relatively long capture on each. I will have a look at how a refocus could fit in when I am poking around in the mosaic code next. I have some changes to make there to try to avoid iOptron software prompting you for 'point at sun?' confirmation every time the mount moves in a solar mosaic too...

cheers,

Robin

[Borodog]

Thanks, Robin.

[Fir Chlis]

If it’s of any value to have a response which doesn’t say anything other than ‘it worked well’, then this is such a report.

Sky was hazed with thin cloud, so unable to see any DSOs for Live Stacking, so I thought I’d have a go at a Lunar Mosaic, as the moon was bigger than my FoV. I didn’t know about the auto-align feature, but it worked like a dream. Two panels of 1000x29ms produced, but I had to wait until the morning to work out how to stack them in Autostakkert, sharpen in Registax, and stitch in GIMP.

I’ll certainly try it again when the Moon is bigger than my FoV.

Geoff

[Borodog]

Robin,

The Moon is coming back so I am trying to be ready to finally get the perfect high res mosaic. I didn't have much luck with any of the alignment methods last month.

This was the automatic method. And this was with the larger FOV ASI183MC, much less my ASI678MC.

lunar_mosaic.jpg (876.4 KiB) Viewed 91 times

There was literally no way to get more of the limb of the Moon in the FOV and it still couldn't find a solution.

The guided method seemed to have too much error. You have to move a point on the limb to 3 different points within a single FOV, but the uncertainty in the calculation seems very large when the FOV is small, because when I slewed to the extents after alignment, they were way off. I'm also not sure how backlash is handled (if at all) during this process; it seems it would be a problem if you don't realize to manually overshoot and return to take up the backlash, always ending with a movement in the same direction, an issue I just thought of, so I definitely wasn't doing it at the time.

With the manual method, it's difficult to assess and set the correct orientation angle from a small FOV, which again leads to a large error at the extents of the mosaic.

These are all problems with the alignment phase, well before the problem I experienced before of drift between panels.

I have been trying to come up with hopefully simple enhancements that could improve this tool. I am just throwing these out there:

1) Global guided alignment. Instead of moving a feature around a single (small) FOV during guided alignment, have the user manually slew to the full extents (requiring movements end in the same directions so backlash is not an issue) and click to set alignment points at the top, bottom, left, right of the Moon (in fact I would use horizontal alignment lines for the top and bottom, vertical alignment lines for the left and right, as when you are using a point it seems difficult to judge whether you are at the correct tangent point). It seems like this would be less sensitive to error.

2) A calibration correction step. After alignment, when you slew to the extents (left, right, top, bottom), let the user correct the location using a similar method to (1).

Thanks,

Mike

[admin]

Hi Mike,

thanks for the feedback - I think I am planning to do most changes to the auto alignment as that tends to be the most accurate (as you noted, the manual only really works for 2 or 3 panel mosaics where you can see a fair bit of the moon to get the alignment right).

Looking at your screenshot, I think the most likely cause of the issues with the auto alignment not being happy with the quality of the edge detection is the low brightness of the moon. My experience has been that a bright moon (even somewhat over-exposed) gives the best results. I have a plan to run the detection twice on each frame - once on the raw image and once on a brightened version if the initial measurement has a problem, but not had time to do that yet. For now, try over-exposing during the alignment stage and see if that helps (if not, just grab a few frames that give a red rather than green circle and send them in my direction).

cheers,

Robin

[Borodog]

I wondered if the brightness of the Moon, which was set to not clip the small, bright highlights, might have had something to do with it. The auto alignment has been able to find a solution in the past with the ASI183MC. It has never, not even once, been able to find a solution with the ASI678MC, however. Do you expect it would be possible? I calculate the best case scenario is that the sensor subtends approximately 38 degrees of arc at the limb:

Automatic_Moon_mosaic.png (345.44 KiB) Viewed 23 times

[admin]

Hi Mike,

I have a simulated camera that draws the current phase of the moon onto an image based on focal length, pixel size, co-ordinates, etc that I can use for testing - in that, it works with a 1920x1080 3.75 micron pixel camera at 2800mm, which is just a bit smaller than the 3840x2160x2 micron of the 678 (I can't change the pixel size as easily as the other stuff, so left it at 3.75 microns). I guess one important difference between that and the real thing is the effects of seeing - they will make your moon edge shimmer, which may cause issues with the threshold for how good the fit to a circle should be.

The actual algorithm to find the offset, rotation and scale is pretty robust, even with only a small portion of the limb in view - it just needs to be able to find the center point of the visible part of the curve in 3 cases.

If you can't get the curve to turn green in spite of boosting exposure, just grab a short video that I can test with - I don't have anything with a 2.8m focal length on hand to test with at the moment.

cheers,

Robin