David's Astronomy Pages Notes - Session 301 (2008-09-11) Notes (S296)   Notes (Main)   Home Page   Notes (S304)

Goto Images from 2008-09-11

Image Linking - Raw Images for Target Centering

It had been noticed over several sessions that there was some underlying issue associated with the realtime 'Linking' of locate images recorded for the purpose of high precision centering of target stars. Particular images containing seemingly adequate number of stars would fail to image link using Insert WCS method on both USNO or GSC catalogs), although they would often link successfuly solve using Fail Safe method.  The impact of the issue was several-fold:

a) annoying delays (up to 8 minutes in total during a 3 hour session)
    - most significantly delays whilst Image Link unsuccessfully tried out numerous alternative Link configurations
    - but also delays whilst extra Locate Images were taken in an attempt to meet high-precision pointing requirements (ie <0.6 arc sec error) 
    - delays whilst additional Locate Images were taken with ever increasing exposure time (eg 10s, 20s, 30s)

b) suboptimally positioned images, where high precision pointing was not achieved, potentially leading to poorly framed images, or unusable images because one or more reference stars lie outside the captured frame.

During S301 session (-15 deg C) a total of least 8 minutes of non-productive time (NPT) can be ascribed to the problem.  Given that session was terminated early due to cloud, the loss of 8 minutes of useful imaging time is equivalent to the loss of image sets for 1-2 target objects.  In comparison the time lost during S294 session (-20deg C) totalled only 2.5 minutes. Whilst this is a smaller loss, it is nevertheless still a very much unwanted loss.

Previous workarounds to mitigate some of the delays associated with failed matches and prevent suboptiminally positioned images have worked quite well, but the large delays associated with failed attempts to Insert WCS remained.

Finally tired of this annoying issue, the problem was thoroughly investigated following session S301.

A review suggested that the problem seemed to effect sessions with warmer CCD temperature (eg  -15 deg C) more so than sessions with colder CCD temperature of -20 or -25 degC.  This has subsequently shown to be due to more numerous /more significant hot pixels associated with warmer images which results in more false stars causing both higher computing overhead and more failed matches during linking attempts.

Raw Image (star field with R CrB) Despite the reasonable number of star objects, the images failed  to link using normal Source Extraction (SExtractor) settings (Note:  the large number of hot pixels in the raw image)

Raw CCD Image 20s, 2x2 binning, C Filter,  2008-09-11 21:40h UT (#301102)

Equivalent star field in TheSky (Software Bisque) 10 stars (with Mag < +15) are available for image linking  The stars can be reasonable easily recognised in the raw image above,  so why should Image Linking fail ?

Raw Image - showing Inventory Stars (CCDSoft, Software Bisque) (Min. number of pixels above threshold = 4,  Detection Threshold = 1.25 ) Note the large number of hot pixels that are picked up as Inventory Stars. These are numerous enough to cause Image linking / Insert WCS to fail.

Raw CCD Image as above

Raw Image - showing Inventory Stars  (CCDSoft, Software Bisque) (Min. number of pixels above threshold = 10,  Detection Threshold = 1.25 ) Note the increasing the min. number of pixels above threshold prevents  the majority of the hot pixels from being picked up as Inventory Stars. This time 'Insert WCS' succeeds with a 9 star solution.

Reduced Image  Dark and Flat Frame correction of above Raw Image Notice that the original hot pixels are largely removed during image reduction

Reduced CCD Image 20s, 2x2 binning, C Filter,  2008-09-11 21:40h UT (#301102)

Reduced Image - showing Inventory Stars  (CCDSoft, Software Bisque) (Min. number of pixels above threshold = 4,  Detection Threshold = 1.25 ) Although a few hot pixels are still picked up as Inventory stars,  'Insert WCS' succeeds with a 12 star solution.

Reduced CCD Image as above

Reduced Image - showing Inventory Stars (CCDSoft, Software Bisque) (Min. number of pixels above threshold = 10,  Detection Threshold = 1.25 ) Few if any hot pixels are picked up as Inventory stars with this less aggressive Star Detection Whilst 'Insert WCS'  does succeed ok it now succeeds with only a 10 star solution  compared to the 12 star solution above.

Reduced CCD Image as above

Link tests based on 46 locate images from Session S301 (-15 deg C). Exposure Times:  5s, 10s, 15s, 20s, 30s, C Filter

Link tests based on 59 locate images from Session S294 (-20 deg C). Exposure Times:  5s, 10s, 15s, 20s, 30s, C Filter

Link tests based on 169  locate images from Session S301 (-15 deg C).   Exposure Times: various. Filters : various

Conclusion is that without image reduction or hot pixel repair, raw images need to be Image Linked using less aggressive star detection if they are to reliably link.  However reduced image can stand more aggressive settings and indeed these are need for more reliable image linking and to pick up the maximum number of solution stars.

A solution is found whereby the Min. number of pixels above threshold value in the Source Extraction set up are set to 10 for realtime observatory work using Raw Images, but set back to 4 for later Image Analysis using Reduced Images.

Back to Top

Image Linking - Reduced Images for DY Per variable star analysis

Monitoring of variable star DY Per has been conducted for the last 9 months using V & R filters.  During this time it had been noticed that R filter images had an unusual tendency to fail to link successfully during Insert WCS operations, even though the images contained lots of stars. Sometimes all 3 R filter images would fail to link, whilst 3 V filter images easily linked even though visually there was little to choose between the V and R filter images.  Whilst it was noticed that the DY Per images did contain an unusually large number of (small) stars the reason behind the images failing to link was a mystery.

After determining a solution to make certain raw images link, by increasing the Min. number of pixels above threshold  from 4 to 10, the same solution was tried out on previous set of 'failed' DY Per images.  The outcome was entirely successful  - previous failed R filter images all successfully linked.

By comparing the magnitude estimates made using V filter images, it was confirmed that the 'Min. number of pixels above threshold ' parameter had no impact on the magnitude estimates and this did not impact photometric result.

Occasional failed links involving R filter images of BL Lac were also fixed by changing 'Min. number of pixels above threshold ' from 4 to 10.

V Filter Image of DY Per (Perseus) This V filter image links with an astrometric solution involving 46 stars

CCD Image 60s, 2x2 binning, V Filter,  2008-02-11 23:08h UT (#262119)

R Filter Image of DY Per (Perseus) This R filter image taken one minute after the above V filter image looks very similar  yet it fails to successfully link. 2 other R filter images taken on the same night also failed to link

CCD Image 45s, 2x2 binning, R Filter,  2008-02-11 23:09h UT (#262120)

Equivalent star field in TheSky (Software Bisque) Around 50 USNO-B stars (with Mag < +15) are available for image linking  The stars can be reasonable easily recognised in the raw image above,  so why should Image Linking fail ?

Raw Image - showing Inventory Stars  (CCDSoft, Software Bisque) (Min. number of pixels above threshold = 4,  Detection Threshold = 1.25 ) Note the large number of faint stars or hot pixels that are picked up as Inventory Stars. These are numerous enough to cause Image linking / Insert WCS to fail.

R Filter CCD Image as above

Raw Image - showing Inventory Stars (Min. number of pixels above threshold = 10,  Detection Threshold = 1.25 ) Note the increasing the min. number of pixels above threshold prevents  many of the faintest stars or hot pixels from being picked up as Inventory Stars. This time 'Insert WCS' successfully succeeds with a 4 9 star solution.

R Filter CCD Image as above

Graphs showing Photometric results from Per DY  including data from newly linked R Filter images

Back to Top