The Astrobin All Sky Survey: A proposal for a community resource Other · Brian Boyle · ... · 363 · 13511 · 59

This topic contains a poll.
Would you be interested in contributing towards an AB all sky survey?
No. I wouldn't find such a survey useful.
No. Satisfactory data already exists for me elsewhere.
No. I would find such a survey useful, but I don't have the time, location or equipment to contribute.
Yes, I would be interested in taking part. One or two fields maximum.
Yes, I would be interested in taking part. Prepared to do multiple fields.
james.tickner 1.20
...
· 
·  1 like
·  Share link
A suggestion for trying to manage gradients with the image stitching. This is just an idea - not something I have to tried at this stage, but I'd be happy to give it a test if someone has some mosaic tiles (ideally with gradients!) that I can test on.

Consider a section of a mosaic shown below, where each lettered region is a 12 x 12 degree field and the darker blue regions are the overlaps. Let us imagine that the fields have been submitted by different users - they have been 'undistorted' using AlignByCoordinates or a similar PI script, and that the user has made their best efforts to remove gradients.

Suppose that we are trying to work on the central field 'E'. Where we have overlaps with the surrounding fields A, B, C etc we calculate the average RGB value for all overlapping images (including the contribution from E itself). Where there is no overlap in the centre of field E we assign a null value. The difference between the RGB values of field E and the average valuesjust calculated provides an estimate of the unwanted gradient of image E in the overlap region. Probably to avoid problems with different star shapes, misaligned stars etc we would remove the stars from all of the images before performing the maths.

We now interpolate this residual gradient value across the centre of tile E. Finally we can subtract the residual gradient to obtain the 'true' image for field E. Repeating this procedure across the grid allows us to make the corrections for the entire sky.

Some comments:

- By construction, the algorithm should eliminate or at least minimise seams between images. Effectively each overlapping image is corrected to have the same RGB values on a scale governed by the interpolation algorithm.
- Whilst the algoithm cannot of course estimate the 'true' gradient perfectly, it makes a best estimate by averaging across the available data taken (presumably) by different observers at different times, maybe using different equipment.
- The algorithm should provide a quantitative approach to QC: any field whose RGB values in the overlap region differ too much is suspect and should be reviewed.

image.png
Like
MichaelRing 4.64
...
· 
·  Share link
When I read correctly the test data from Brian had some gradients, it is still available for download:

https://www.mycloud.ch/s/S00B9DF137C56385DEA23517B17090B96F38766499E

Michael
Like
james.tickner 1.20
...
· 
·  Share link
Michael Ring:
When I read correctly the test data from Brian had some gradients, it is still available for download:

https://www.mycloud.ch/s/S00B9DF137C56385DEA23517B17090B96F38766499E

Michael

Thanks! I'll take a look
Like
MichaelRing 4.64
...
· 
·  Share link
I spent some time today to come up with an efficient workflow for preparing the observations (in Nina) so for starters, here is a proposal (together with some questions) on how something efficient could look like. I also checked Astrobin's equipment database for popular lenses in the range of 80-250mm, the most relevant results were the Redcat (250mm) followed by Samyang/Rokinon (135mm) Askar FMA180 (180mm) and Askar ACL200 (200mm) and some more lenses with 135mm FL. More on that later in this post....

So first, let's start with preparing our field, for this I used the Catalog Star Generator Script in Script->Render

Bildschirmfoto 2023-06-01 um 15.57.52.png

  I entered RA/DEC from one of the tiles from the Excel Sheet, set the Observation Date to today, Rotation to 0° and then entered focal distance (135mm) and the Pixel Size of the camera I wanted to try (Nikon Z9 as an example for a full frame Camera).
Then I needed to calculate the size of the generated picture. We want 12° wide tiles, so we need to do the math: 12*60*60 = 43200 Arcseconds are required and our camera has a resolution of 6.56992 Arcseconds per pixel.
So we need to divide 43200 by 6.56992 to get the pixel with of the area that covers the 12°: 43200 / 6.56992 = 6576 pixel
You now enter this number in the Dimensions fields.

I then had to reduce the Maximum Magnitude to be able to run the script, after setting it to 14 things went through smoothly and I saw the freshly created StarMap in PI.

QUESTION: You guru's out there, what would be the best projection? This question was not yet clarified, unless I missed something.

Next Step is now to save the Starmap to an xisf file, then our work in PI is done and we continue in the framing Wizzard of Nina:

Bildschirmfoto 2023-06-01 um 16.12.15.png

To load the Starfield, choose Image Source: File
Now make sure that the data for Width/Height/Pixel Size/Focallength matches what you entered for the Starfield and that Rotation is 0°
Then click the Load Image button and Nina will fill out RA/DEC and the field of view for you after analyzing the file.
Now make sure that the data for Width/Height/Pixel Size/Focallength matches your equipment and that Rotation is 0°

As you can see, the combination of a full Frame Camera with a 135mm lens nearly can do the full tile in one session.

QUESTION: Why did we choose an aspect ratio of 1:1 for the tiles, is there a technical reason? Why not something like 3:2 that better matches the sensors of our cameras in the regions with many tiles to shoot? Perhaps there is a good reason, I simply do not know and think there could be a bit of optimization, especially when I go to the case of a APS-C sized sensor:

Bildschirmfoto 2023-06-01 um 16.26.06.png

When you look at the 4 tiles, I am not able to cover the full required field with 4 Panels, something is missing on top center and in the bottom edges.
Yes, I can switch 'Preserve alignment' switch on, then everything will fit, but I would have to re-adjust the camera angle after each pane, which is a no go when you care for your night's sleep or have other scopes to attend in your nightly session.

By choosing a 3:2 aspect we can go up to 70° Dec and could still have a complete tile with a width of 12°.

One could argue that you simply need to book more tiles to be able to produce 12x12° in the end, but there is a big risk involved because you can only submit your first correct tile after doing 6x2-4hrs of integration and because our rules for the moon are quite strict you will need more than one cycle to finish the data, which might cause gaps in the pano because you cannot make it in time because of bad weather.

Penny for your thoughts....

I will put the discussion about if it makes sense to include RedCats in the came to a later post, have to run.

Michael
Edited ...
Like
james.tickner 1.20
...
· 
·  1 like
·  Share link
@Michael Ring Good observation about the convenient tiling of the field. @Brian Boyle might have a comment on the original selection of the 12 x 12 field size. If a different non-square size is more efficient it is straightforward for me to regenerate the tile centre model.
Like
andreatax 9.89
...
· 
·  1 like
·  Share link
I concur that the square tile doesn't fit most sensors out there (including the one I'm thinking of using) so we should really drop it. I'd also reduce the SNR from 30 to something smaller as there are far too many tiles and not enough poeple. As for the projection model it wasn't discussed so far. Marcatore would do for me or we can do a spherical mapping on a 3D (WRML/X3D) model, that would be fun…
Like
MichaelRing 4.64
...
· 
·  Share link
The (from my point of view) good news is that when we set aspect ratio to 3:2 also RedCat Users with Full Frame Cameras can join the party:Bildschirmfoto 2023-06-01 um 18.21.02.png

But they should not go higher in DEC than 30-40, otherwise there will be parts missing.
But with Preserve alignment enabled they can cover the full field with only 4 exposures in the full DEC range.

Michael
Like
profbriannz 17.56
Topic starter
...
· 
·  Share link
Great discussion everyone.  I think we should change field size/aspect ratio. 

My original rationale was simply to avoid enforcing a given PA on individuals - but I agree it is not an efficient use of sensor field.

Given all the work that has been done, I am also happy to review field-to-field spacing and overlap.  My original suggestion was motivated by a compromise over the range 85-200mm focal lengrh.  

At 200mm a full frame sensor is 10.3 x 6.9 deg

135mm  is 15.3 x 10.2

85mm is 24.3 x 16.8

So fields on 15 x 10 deg centres would allow these lenses to cover the area in

200mm:  4 panes  18x12 degrees with 20overlap 

135mm: 2 panes  18x 15 degrees with two panes

85mm 1 panes 24 x 17 degrees

I have no doubt someone could come up with a somewhat more efficient tiling strategy.  I just chose 10 x 15 because they were both divisible into 180… 

Of course this assumes full frame sensor.  For APSC it will require more panes per field centre.  We could drop field spacing back to, say, 9 x 6 to accomodate APSC.

More than double the number of fields - but those with FF sensors wouldn’t need to tile (although the overlaps between fields given by 200mm lenses would be quite low.  



CS Brian
Like
profbriannz 17.56
Topic starter
...
· 
·  Share link
Michael Ring:
Al little training data will be nice, if you want you can put the data here:

https://www.mycloud.ch/s/S00B9DF137C56385DEA23517B17090B96F38766499E

Data will be automagically deleted after 30 days, until then whoever wants to play with the data can access it via the link.

Michael

*** Type your reply here **

@Michael Ring

I ran WBPP again on 4 x panes for the "old" survey Field 70, but this time stopped at WBPP - so no ABE, SPCC, BXT or NXT.  I can upload the .xisf  files to the site if people want to experiment.  

Brian
Edited ...
Like
MichaelRing 4.64
...
· 
·  Share link
I have enabled uploads on the share again, if you cannot delete files then do not bother, I will clean up after you tell me that uploads are done.

And, I tend to disagree with your tile size, I did not check 200mm (but those lenses are rare compared to 135mm) and with 12°x8° we have nice framing for 135 FF 135 APS-C and 250mm FF, perhaps you could check how 200mm FF behaves with this tile size, I will try later, I am currently preparing my new Samyang 135 for it's task by running aberatiation insptor and tuning the backfocus and the clouds are supposed to roll in any  minute….
Like
profbriannz 17.56
Topic starter
...
· 
·  Share link
@Michael Ring

Thanks for that.  File upload initiated.  

Happy to adopt 12 x 8 as field centres.  Would take 3 x 200mm panes at 20% overlap giving 17 x 10 coverage for each 12 x 8 field.  Could drop overlap to 10% to give 13 x 10 coverage, but I think this is too tight.  If we adopted this field size, then 200mm users (including me) might be tempted to do stripes along long axis to cover range of fields.

For this reason, I might suggest that we orient field centres with RA being the long axis.

Brian
Like
GoldfieldAstro 0.90
...
· 
·  Share link
One other thought I’ve had, all of the tiles should be rotated to 90°. If you’ve got a DSLR that you’re attaching to a mount it’ll likely be through the bottom bolt of the DSLR if the lens doesn’t have a rotatable tripod mount. When attached in this fashion the sensor is either at 90° or 270°.
Like
MichaelRing 4.64
...
· 
·  Share link
Hmmm... I thought that you started with the 12° tiles so that there is enough flesh to actually do stitching with 10% overlap of 10° tiles (which makes a lot of sense to me) , so your 13x10 coverage with 2 frames can work, at least for me in Nina with 15% overlap things with FF and 200mm look fine at 50° DEC

Bildschirmfoto 2023-06-02 um 01.54.05.png
Like
andreatax 9.89
...
· 
·  Share link
I wonder how many of those interested in this project have a FF camera up to the task. I certainly don't.
Like
MichaelRing 4.64
...
· 
·  Share link
This is why I also did calculations for APS-C, I also do not own a Full Frame Camera that is Astro Modded.
For APS-C to work in an efficient way you will either need to own a 135mm lens or you should think about partnering with another APS-C User to jointly do a tile.

Rotation inside of the tile should from my point of view be left to the user to find the best possible match for the own equipment. I for example usually have L-Brackets on my DSLR cameras so that I could choose between mounting the camera in 0° or 90° position. (But as said, none of my DSLR's is Astro modded, i will use an 'old' ZWO ASI 071 camera for this project and have 3D printed the necessary tube rings for mounting the lens.
Edited ...
Like
profbriannz 17.56
Topic starter
...
· 
·  Share link
Agree that rotation should be left to user, but if we go to non-square fields, then then field-centre distribution has to be fixed in one orientation.  Again this should be decided on the principle of inclusivity, if there is a field centre distribution that excludes a sizeable part of the images.  A 12 (RA) x 8 (Dec) would work for me - since I could do Dec strips with the long axis of my sensor NS.  The other way round would be a pain (not pane) as RA strips would be harder to do, since the RA field boundaries shift with dec.  

Note to that my original proposition was based around 135mm and FF.  I proposed it not because I owned it (I have a 200mm/FF), but because I thought this was the more common set-up.

However, I am convinced by @Michael Ring argument that 12 x 8 field centres works best for most.  Although this will require 4 panes per field for an APS-C sensor at 135mm.

The other option might be to go for a smaller field size (at the risk of increasing the number of fields) - say 9 x 6 deg field centres.  So this could be done in 2 panes with 135mm/APS-C.  And one shot with FF for all lenses up to 200mm [at the risk of inefficiencies]

@andrea tasselli  Are you saying that 12 x 8 (ie four panes per field) is too much for you and others with APS-C?  (I do sympathise with this!)  Would you rather have smaller field-to-field centres?

From the principle of inclusivity, I think we would not want to make it difficult for APS-C users.   Smaller field centres also help individuals with FF sensors have most flexibility over observing, since now it is only one pane per field. 

Brian
Like
james.tickner 1.20
...
· 
·  Share link
Just to confirm, are we currently looking at 12 x 8 fields with 20% overlap ie 10 x 6.4 degree centres approx? And with long axis of rectangle running parallel to RA? I can rerun the field centre calcs this evening when I get home to check the number of tiles needed.
Like
profbriannz 17.56
Topic starter
...
· 
·  Share link
@James Tickner I thought it was 12 x 8 field centre spacing, but I may have misunderstood.  Since @Michael Ring has done the modelling, perhaps he can confirm.  9 x 6 field-to-field  (10.5 x 7 min field size ) would simplify things a lot and make it easy for APS-C owners.

CS Brian
Like
andreatax 9.89
...
· 
·  Share link
Brian Boyle:
@andrea tasselli Are you saying that 12 x 8 (ie four panes per field) is too much for you and others with APS-C? (I do sympathise with this!) Would you rather have smaller field-to-field centres?


Smaller fields would certainly help, as trying to cover a larger field with a smaller sensor introduces additional wastage. APS-C is 2/3 of a FF so it may be possible to have two tilings, one with FF sensor and one with APS-C and each field comprising either 2 FF tiles or 3 APS-C tiles.

It would also be interesting to know the potential partecipiants' hardware availability and locations, so we would known what is what and where is distributed. Maybe using Google Forms?
Like
james.tickner 1.20
...
· 
·  Share link
I've put the various combinations into a spreadsheet (see below). summarising tiling result for three different field sizes (6 x 9. 10 x 12 or 10 x 15), three different focal lengths (80, 135 or 200) and two different sensor sizes (APS-C or full frame). I also had a look at 12 x 12 field but this really doesn't work well for any sensor size of focal length combination.

The 'N' column shows the number of images required to complete the field, and the % waste is the percentage of collected area that is not required (note that this simple calculation ignores overlaps required between images that make up the field, so with N >= 2 we really want a waste factor of 10% or more). N values less than one indicate that 2 (N = 0.5) or 4 (N = 0.25) neigbouring fields could be covered by one image.

We see that 6 x 9 and 10 x 12 fields are good for APS-C/135 mm and FF/135 or 200 mm combinations, but rather wasteful for others. The 10 x 15 field is good for APS-C/80 or 135 mm, and FF/80, 135 or 200 mm. So overall the 10 x 15 degree field might be the best choice to support a wide range of camera and lens combinations with best efficiency. 

image.png
Like
Astrogerdt 0.90
...
· 
·  Share link
I just updated the processing guide to the third draft to include the distortion correction using AlignByCoordinates script. Unfortunately, this added a little complexity because doing it after the integration would result in a loss of SNR, so it had to be done prior to integration. The procedure is described here: https://docs.google.com/document/d/1QZxLRpfVuxSxTDWNpjHYDpZ7p9Ua6vdz/edit?usp=sharing&ouid=102793495713995642568&rtpof=true&sd=true

Changes can be found in the "Preperation for WBPP" and "Preprocessing in WBPP" -> "Registration Reference Image". 

Is that added complexity OK for everyone?

CS Gerrit
Like
andreatax 9.89
...
· 
·  Share link
I just updated the processing guide to the third draft to include the distortion correction using AlignByCoordinates script. Unfortunately, this added a little complexity because doing it after the integration would result in a loss of SNR, so it had to be done prior to integration. The procedure is described here: https://docs.google.com/document/d/1QZxLRpfVuxSxTDWNpjHYDpZ7p9Ua6vdz/edit?usp=sharing&ouid=102793495713995642568&rtpof=true&sd=true

Changes can be found in the "Preperation for WBPP" and "Preprocessing in WBPP" -> "Registration Reference Image". 

Is that added complexity OK for everyone?

CS Gerrit

I don't use WBPP so I can't comment on the written procedure but as far as I know if you register images using the "Thin Plast Splines" option with "Distortion Correction"  and "Local Distortion" selected than the registered integrated light should be undistorted.
Like
MichaelRing 4.64
...
· 
·  Share link
Doing more calculations with smaller tiles and just found another can of worms, APS-C is not APS-C....

Today I remembered that Canon Cameras always have a 1.6 factor instead of 1.5 for Nikon. So I once again checked the Astrobin Equipment Database (I LOVE this feature of Astrobin....) and looked at sensor sizes of the most relevant Canon APS-C Cameras and their size is (worst Case) 22.2x14.8 with a resolution of 3888x2592 5.7um pixel size meaning that the fields covered by this camera are:

85mm: 14.9x9.96
135mm: 9.4x6.27
200mm: 6.34x4.23
250mm: 5.07x3.39

Here, for comparison the field sizes of FF (taken from a Canon 6D):
85mm: 24.15x16.1
135mm: 15.21x10.14
200mm: 10.27x6.84
250mm: 8.21x5.48

When looking at possible combinations I gave preference to 135 and 250mm as those are the most used FL (Samyang/Rokinon and RedCat)
and came up with:

10x6.66 (with an inefficiency disadvantage for 135mm APSC), so I did not go into detail)
and
9x6 which seems to produce acceptable matches for most combinations so going into more detail:

APS-C 135mm: 1 Tile (Rotation 0°)
APS-C 200mm: 3 Tiles (Rotation 90°) 4 Tiles (Rotation 0°)
APS-C 250mm: 4 Tiles (Rotation 0°, works up to 50°) 6 Tiles (Rotation 90°)

FF 135mm: Can cover 2 Fields in 1 shot
FF 200mm: 1 Tile (Rotation 0°)
FF 250mm: 4 Tiles (Rotation 0°)

I would not go lower than this as the number of required fields will already by now have exploded, @James Tickner , would you mind doing a calculation based on a Field to Field Grid of 7.5x5 (based on 9x6 min Field size)

I also re-did a bigger size, because of the Canon APSC Image size I reduced to 11x7.3, 20% overlap:

APS-C 135mm: 2 Tiles (Rotation 90°) 4 Tiles (Rotation 0°)
APS-C 200mm: 4 Tiles (Rotation 0°)
APS-C 250mm: 8 Tiles (Rotation 90°) 9 Tiles (Rotation 0°)

FF 135mm: 1 Tile
FF 200mm: 2 Tiles (Rotation 90°) 4 Tiles (Rotation 0°)
FF 250mm: 3 Tiles (Rotation 90°) 4 Tiles (Rotation 0°)

 Again here, @James Tickner a calculation of the number of required tiles would be helpfull, 9.1x6 Field to Field Grid based on 11x7.3 min Field Size


Hope I did not do anything wrong with my calculations and my thinking, is 10% overlap between 2 completed fields enough?

I am still wrapping my head arround what @Brian Boyle wrote:

Quote: I might suggest that we orient field centres with RA being the long axis

Is this what you expect (I annotated the image so that RA/DEC are visible in the frame):

Bildschirmfoto 2023-06-02 um 11.19.09.png

The more I think about it the more this looks wrong, you want the fields to be rotated by 90 degrees like this, correct?

Bildschirmfoto 2023-06-02 um 17.48.51.png
Edited ...
Like
MichaelRing 4.64
...
· 
·  Share link
@James Tickner : sorry, did not see your post until posting mine, I wrapped my head arround this topic for the last few hours (good that I have vacation atm and the better half is working..), I think it makes sense to include 250mm for all the RedCat's out there....

And I could not find a relevant number of 85mm lenses in the Astrobin DB so I ignored them...

Michael
Edited ...
Like
Astrogerdt 0.90
...
· 
·  Share link
@andrea tasselli The distortion correction does only correct distortion relative to the reference frame, not on an objective scale in comparison to the coordinate grid. AlignByCoordinates removes the distortion from the image by aligning it in a way that it matches the coordinate grid, and then StarAlignment aligns every single image to that reference using DistortionCorrection, so that there are no alignment errors in the corners. 

CS Gerrit
Like
 
Register or login to create to post a reply.