CCD Superintendent

[MarkS]

I like the concept of estimating collimation from FWHMs of stars in a starfield but there are are number of things I don't like about CCD Inspector:
1) It gives no feedback to show which stars it used
2) It gives no information on the amount of, or the direction of, star elongation.  Actually, to be fair, CCD Inspector 2 will provide an "aspect ratio" plot which gives an idea of the amount of elongation but not the direction.  The directions of elongation of the various stars provides some very useful information.  For instance, sometimes the elongations tend to "radiate" from a fixed point (the optical centre), sometimes they are tangential instead of radial and sometimes they are all in the same direction (because of tilt or becuase of guiding issues).

Well today I've tackled the maths (fitting centroids and moments of inertia) and have written some code to calculate the direction of the major and minor axes of an elongated star and to calculate the FHWM along these axes (note that this is different to calculating the FHWM in the x direction and the y direction, which is something that IRIS does even though it is not very useful).

So I'm now well on the way to being able to identify all stars in an astro-image and to calculate the direction of elongation of each star together with their FHWMs along the star's major and minor axes.  So the next question is how to graphically display all this data, ideally all in a single plot. Any ideas gratefully received!

Mark

[Mac]

Nice topic,

Would it be possible to show the stars as a 3d plot,
Assume the following.
your optics are perfect, i.e. in collimation ect,ect.

then if you plotted your stars as 3d isometric (or what ever its called)

you could display them as a grey scale cone upto a peak.

Now if your collimation is off, then the cone would not be a true cone,
you could overlay a red and green cone to show how much your optics are out by
in the x/y direction.
if that make sense.

ill try and explain that a bit better.

looking from above you would get a perfect cone,



if you collimation was out to the left you could show the overlay in a different colour,


the same for top to bottom,


and then all three,


but if they were shown as cones,

then as you adjust then the cones would converge onto the perfect grey one.

something like this.  (i know the red is on the other side now, but you get the idea)


The joys of solidworks.


Mac

[RobertM]

Not sure if this is helpful but I tend to look at two things but CCD Inspector only provides one.

1) It does analyse images as they are downloaded through MaximDL and gives a fair representation of the direction and how far out collimation is.  However, the 'live' display seems to be very susceptable to seeing so even when left to settle the collimation point does bounce too much to collimate with any degree of accuracy.  At the moment the only way around this is to take long exposures but maybe an 'averaging' of close (ignoring outlier) results would be better.

Mac's idea seems good though I would add that the circles could be indicative of the 'criticality zone' so when they overlap then collimation is 'close enough'.  For fast optics the circles might be small and slower scopes slightly larger - just a thought

2) When I take an image I always look at the corner stars as these are the most affected by collimation errors.  It would be really good to have, in addition to an indication of the error of the centre of collimation, a display of the corners so the actual star shapes can be seen - I see this as the acid test.

Sounds like an excellent project - hope it won't be just for DSLR's 

Robert

[MarkS]

Thanks for the suggestions - I haven't yet got to the stage where I will plot the resulting data.  Although I'm making some good progress with this, I'm hitting a few obstacles.

I've reached the stage where my program will look at an image, identify the stars (this was the most difficult part) and then print a complete list of centroids, star size (in pixels), total intensity, angle of elongation and FWHM along the major and minor axes of elongation.

Analysing this data has raised a few issues:
1) There is a strong correlation between star size and calculated FWHM - the larger the star, the greater the FWHM.  I think this is due to 2 reasons:
a) Smaller stars are dimmer and so the outer edges are more difficult to accurately distinguish from the background noise - I'm probably erring too far in the direction of avoiding false positives
b) Larger stars are brighter and closer to saturation which makes them wider and flatter
Interestingly, if I analyse star data from IRIS I see exactly the same correlation.

2) For stars taken individually, the angle of elongation is not terribly reliable.  However, plotting the angles for stars in a certain vicinity does show some quite strong "bunching" of angles.  I'm beginning to think that some kind of wind rose plot might be the best approach i.e. split the image into regions and display a wind rose for each region.

The general approach I am taking is to identify a star, calculate the backgound level and standard deviation in the area around the star, identify the pixels comprising the star and then doing the centroid calculations (analysis of moments up to second order) on those pixels once the background is subtracted.

However, the centroid calculations are easily upset by wrongly including/excluding any individual pixel or by having an incorrect estimate of the local background flux.  It may end up being more fruitful to take a curve fitting approach (i.e. calculate the parameters of a best fit 2D Gaussian).  Maybe this would help solve the first issue.

Anyway, it's all interesting stuff and preferable to hanging out on street corners drinking cider

Mark

[RobertM]

I'm beginning to think that some kind of wind rose plot might be the best approach

That's a novel approach and I think it could work quite well considering the variability of the data.  Several images could be averaged if a more accurate result was needed or the starfield was less then ideal.

Would this work on fits files written to disk or had you planned some other 'live' feed - sorry for getting ahead here I can see this approach working very well for the Hyperstar

Robert

[MarkS]

Would this work on fits files written to disk or had you planned some other 'live' feed - sorry for getting ahead here I can see this approach working very well for the Hyperstar

At present I can open only TIFF files.   I intend to include RAW file support using Dave Coffin's DCRAW so it will work for Canon's raw files.  So any file format supported by DCRAW should also work.  I've also investigated how to code up functionality to monitor a folder for new files - that shouldn't be difficult to implement.  So as long as image files can be dumped into the monitored folder in a supported format, they can be automatically opened and processed.

Potentially I could also add FITS support if it's not already supported by DCRAW.

[Mac]

How about a maxim plugin,
I know there are a few bits of software out there that monitor the image capture from maxim,
rather then doing their own
so that might kill a few birds with one stone, by just accessing the image pipe from maxim.

Mac.

[MarkS]

How about a maxim plugin ...

It might be possible but, at first, I want something completely independent of other software.

Mark

[Rocket Pooch]

Mark,

You need to plot each star centroid and an angle of center movement against an assumed central point in the frame, this will give you the central off set and also allow you to calculate relative collimation against each star.  You will also need to ignore saturated stars.

I'll get a piece of paper and draw you a diag when I get 5.

Chris

[MarkS]

You need to plot each star centroid and an angle of center movement against an assumed central point in the frame, this will give you the central off set and also allow you to calculate relative collimation against each star. 

Not quite sure what you mean but suggestions are welcome.  I might totally ignore them of course 

Mark

[MarkS]

Would this work on fits files written to disk or had you planned some other 'live' feed - sorry for getting ahead here I can see this approach working very well for the Hyperstar

Potentially I could also add FITS support if it's not already supported by DCRAW.

Robert,

If the application could automatically pick up and process any FITS file that appears in a chosen folder, would that work for you?

I've got the "5 pane" view working - it begins by displaying the centre and the extreme corners of the image but afterwards, it is possible to "slide around" each image section within each pane independently.  It does an auto log-scaling on the intensity, for display purposes.



Jumping ahead a bit, I would love to be able to select a bright star in each pane and display the Bhatinov focusing accuracy for each of the 5 stars simultaneously.  This might even be useful during collimation - I'm not sure.

Mark

[RobertM]

If the application could automatically pick up and process any FITS file that appears in a chosen folder, would that work for you?

Yep that sounds good.

I've got the "5 pane" view working - it begins by displaying the centre and the extreme corners of the image but afterwards, it is possible to "slide around" each image section within each pane independently.  It does an auto log-scaling on the intensity, for display purposes.

Nice ! One for the future - maybe you could consider a zoom in/out feature that affected all windows simultaneously.

Jumping ahead a bit, I would love to be able to select a bright star in each pane and display the Bhatinov focusing accuracy for each of the 5 stars simultaneously.  This might even be useful during collimation - I'm not sure.

You could calculate the relative Bhatinov error depending on distance from the image centre which might give a more critical collimation result - having to refocus each time would be a real pain though possibly a necessary step.

All this is great for Hyperstar and other systems with a single collimation control but what about Newtonians ? Would this help you with the E180 ?

Robert

[MarkS]

All this is great for Hyperstar and other systems with a single collimation control but what about Newtonians? Would this help you with the E180 ?

It will certainly help getting the two-part Canon adapter ring correctly adjusted on the E180. Apart from that it would be useful as a critical focusing tool for both the Celestron C11 and the Tak E180.

[MarkS]

A possible way of displaying stars selected for the FWHM calculations with their calculated orientations.  Note that the saturated stars are ignored because their FWHMs would be misleading.

[mickw]

If the green lines are the direction of elongation, there a some weird occurances of adjacent or overlapping stars with perpendicular elongation.

Could that be caused by imperfections with the mirrors or gas/dust in the area of the subject ?