Skip to end of metadata
Go to start of metadata

Links to Astrometry and WCS References:

More information on WCS than you ever wanted to know.

 


It's been surprisingly difficult to get astrometry.net to latch onto the images we produce. This is presumably due to the wide field and distortion. 

For one object catalog I did get a solution, which indicates the center of the image is pointed at dec of -28.7. This was for image 

ut011314.0360.long.fits

which gave 

RA,Dec = (109.299,-28.7017), pixel scale 169.598 arcsec/pix.

here is the corresponding wcs.fits file:

wcs.fits

FITS images are N up E left!

and here it is listed out:

cp65239:Downloads cstubbs$ imhead wcs.fits

SIMPLE  =                    T / Standard FITS file

BITPIX  =                    8 / ASCII or bytes array

NAXIS   =                    0 / Minimal header

EXTEND  =                    T / There may be FITS ext

WCSAXES =                    2 / no comment

CTYPE1  = 'RA---TAN-SIP' / TAN (gnomic) projection + SIP distortions

CTYPE2  = 'DEC--TAN-SIP' / TAN (gnomic) projection + SIP distortions

EQUINOX =               2000.0 / Equatorial coordinates definition (yr)

LONPOLE =                180.0 / no comment

LATPOLE =                  0.0 / no comment

CRVAL1  =        109.320631076 / RA  of reference point

CRVAL2  =       -28.6668823198 / DEC of reference point

CRPIX1  =                 1192 / X reference pixel

CRPIX2  =                960.5 / Y reference pixel

CUNIT1  = 'deg     ' / X pixel scale units

CUNIT2  = 'deg     ' / Y pixel scale units

CD1_1   =      -0.047513171544 / Transformation matrix

CD1_2   =    -0.00146819731348 / no comment

CD2_1   =   -0.000864857162952 / no comment

CD2_2   =      0.0466847846798 / no comment

IMAGEW  =                 2382 / Image width,  in pixels.

IMAGEH  =                 1919 / Image height, in pixels.

A_ORDER =                    2 / Polynomial order, axis 1

A_0_2   =     4.6078119764E-06 / no comment

A_1_1   =    1.23847648274E-05 / no comment

A_2_0   =   -5.29280627903E-06 / no comment

B_ORDER =                    2 / Polynomial order, axis 2

B_0_2   =    4.65369941262E-06 / no comment

B_1_1   =    2.97633768011E-05 / no comment

B_2_0   =   -5.10163633601E-05 / no comment

AP_ORDER=                    2 / Inv polynomial order, axis 1

AP_0_1  =    -9.0981000082E-05 / no comment

AP_0_2  =   -4.58575725434E-06 / no comment

AP_1_0  =   -0.000380322057888 / no comment

AP_1_1  =   -1.23836775631E-05 / no comment

AP_2_0  =    5.29784239136E-06 / no comment

BP_ORDER=                    2 / Inv polynomial order, axis 2

BP_0_1  =   -0.000401626462877 / no comment

BP_0_2  =   -4.59507257312E-06 / no comment

BP_1_0  =   -0.000786806957891 / no comment

BP_1_1  =   -2.97521982273E-05 / no comment

BP_2_0  =    5.09931822201E-05 / no comment

HISTORY Created by the Astrometry.net suite.

HISTORY For more details, see http://astrometry.net .

HISTORY Subversion URL

HISTORY   http://astrometry.net/svn/tags/nova/2013-05-24-2/util/

HISTORY Subversion revision 22686

HISTORY Subversion date 2013-05-06 22:33:22 +0000 (Mon, 06 May

HISTORY   2013)

HISTORY This WCS header was created by the program "blind".

DATE    = '2014-01-14T04:03:54' / Date this file was created.

COMMENT -- blind solver parameters: --

COMMENT Index(0): /data1/INDEXES/200-current/index-219.fits

COMMENT Index(1): /data1/INDEXES/200-current/index-218.fits

COMMENT Index(2): /data1/INDEXES/200-current/index-217.fits

COMMENT Index(3): /data1/INDEXES/200-current/index-216.fits

COMMENT Index(4): /data1/INDEXES/200-current/index-215.fits

COMMENT Index(5): /data1/INDEXES/200-current/index-214.fits

COMMENT Index(6): /data1/INDEXES/200-current/index-213.fits

COMMENT Index(7): /data1/INDEXES/200-current/index-212.fits

COMMENT Index(8): /data1/INDEXES/200-current/index-211.fits

COMMENT Index(9): /data1/INDEXES/200-current/index-210.fits

COMMENT Index(10): /data1/INDEXES/200-current/index-209.fits

COMMENT Index(11): /data1/INDEXES/4100/index-4119.fits

COMMENT Index(12): /data1/INDEXES/4100/index-4118.fits

COMMENT Index(13): /data1/INDEXES/4100/index-4117.fits

COMMENT Index(14): /data1/INDEXES/4100/index-4116.fits

COMMENT Index(15): /data1/INDEXES/4100/index-4115.fits

COMMENT Index(16): /data1/INDEXES/4100/index-4114.fits

COMMENT Index(17): /data1/INDEXES/4100/index-4113.fits

COMMENT Index(18): /data1/INDEXES/4100/index-4112.fits

COMMENT Index(19): /data1/INDEXES/4100/index-4111.fits

COMMENT Index(20): /data1/INDEXES/4100/index-4110.fits

COMMENT Index(21): /data1/INDEXES/4100/index-4109.fits

COMMENT Field name: job.axy

COMMENT Field scale lower: 15.1134 arcsec/pixel

COMMENT Field scale upper: 272.04 arcsec/pixel

COMMENT X col name: X

COMMENT Y col name: Y

COMMENT Start obj: 0

COMMENT End obj: 0

COMMENT Solved_in: (null)

COMMENT Solved_out: (null)

COMMENT Solvedserver: (null)

COMMENT Parity: 2

COMMENT Codetol: 0.01

COMMENT Verify pixels: 1 pix

COMMENT Maxquads: 0

COMMENT Maxmatches: 0

COMMENT Cpu limit: 600.000000 s

COMMENT Time limit: 0 s

COMMENT Total time limit: 0 s

COMMENT Total CPU limit: 0.000000 s

COMMENT Tweak: yes

COMMENT Tweak AB order: 2

COMMENT Tweak ABP order: 2

COMMENT --

COMMENT -- properties of the matching quad: --

COMMENT index id: 4116

COMMENT index healpix: -1

COMMENT index hpnside: 0

COMMENT log odds: 1.6833

COMMENT odds: 5.38327

COMMENT quadno: 2848

COMMENT stars: 3597,3254,3606,0

COMMENT field: 67,52,32,0

COMMENT code error: 0.00321729

COMMENT nmatch: 4

COMMENT nconflict: 1

COMMENT nfield: 200

COMMENT nindex: 820

COMMENT scale: 169.598 arcsec/pix

COMMENT parity: 0

COMMENT quads tried: 3410976

COMMENT quads matched: 3383383

COMMENT quads verified: 156020

COMMENT objs tried: 68

COMMENT cpu time: 67.2642

COMMENT --

END

but that might be enough to get us started....:

 

cp65239:Downloads cstubbs$ xy2sky -d wcs.fits 1000 1000

119.38013 -26.40007 J2000 1000.000 1000.000 

SO let's take the SIMBAD catalog and transform into coordinates for this image.

526  sky2xy 360.wcs.fits @Simbad.radec.dat | grep -v "off" > Simbad.xy.dat

  527  more Simbad.xy.dat 

  528  awk '{print $5, $6}' Simbad.xy.dat 

  529  awk '{print $5, $6}' Simbad.xy.dat > xy.dat

  530  ds9mark.pl ut011314.0806.long.fits xy.dat

Yuk. 

Seems almost OK at the center but rapidly degrades as you move out. 

image was taken at UT Jan 14 03:39 which, according to iObserve, corresponds to LMST=06:30 or in decimal degrees it's 97.5 degrees. That's a little off from the nominal center of the image being 109.35, but maybe it isn't pointed straight up? 

Tried astrometry.net with initial center guess of 97.5 and -30, with third order polynomial fit. It nailed it instantly!

RA of 96.00 and dec of -28.354

Upload Settings
Parity:try both simultaneously
Scale Units:arcseconds per pixel
Scale Type:bounds
Scale Lower Bound:150.0
Scale Upper Bound:190.0
Star Positional Error:5.0 pixels
Limits:RA, Dec - (97.5, -30.0)
radius - 10.0
Downsample Factor:2
Tweak Order:3
Source detection:SExtractor
CRPIX:center

 

It does seem the field center is correct, but there are bad mis-matches between the Simbad catalog and the objects in the field, gets worse away from the center. 

Try with out own photometric catalog, and matching to a subsequent image. 

 

get image 460:

cp65239:calib cstubbs$ scp christopherstubbs@139.229.19.220:/Volumes/2TB/ut011314/ut011314.0460.long.cr2.gz .

cp65239:calib cstubbs$ cr2fits -bw ut011314.0460.long.cr2

DATE-OBS= '2014-01-14T04:41:24' // Exposure date

so this is one hour and two minutes later than image 360, which should add 15.5 degrees in RA, or 111.5 RA and -28.354 in dec. Try astrometry.net for that one. 

Yup, nailed that one too. This is not so bad. 

Center (RA, Dec):(111.372, -28.307)
Center (RA, hms):07h 25m 29.341s
Center (Dec, dms):-28° 18' 25.392"
Size:135 x 89.9 deg
Radius:81.055 deg
Pixel scale:168 arcsec/pixel

 

So let's take photometry catalog from image 360 and transform into pixel coords of image 460.

  579  sex ut011314.0360.long.fits

  580  more test.cat

  581  grep -v '#' test.cat | awk '{print $9, $10}' 

  582  grep -v '#' test.cat | awk '{print $9, $10}' > 360.xy.cat

  583  xy2sky image.360.wcs.fits @360.xy.cat

  584  xy2sky -d image.360.wcs.fits @360.xy.cat

 587  xy2sky -d image.360.wcs.fits @360.xy.cat | awk '{print $1, $2}' > 360.radec.cat

  588  mv ~/Downloads/new-image.fits ./image.460.wcs.fits

  589  sky2xy image.460.wcs.fits @360.radec.cat

  590  sky2xy image.460.wcs.fits @360.radec.cat | grep -v "off" | awk '{print $5, $6}'

  591  sky2xy image.460.wcs.fits @360.radec.cat | grep -v "off" | awk '{print $5, $6}' > 460.xy.cat

The stars do line up in the center, but there is a substantial mismatch away from the image center. 

The offsets here are from one hour of sky rotation times the distortion map error, and are large!

So we need a better distortion map, or initial WCS solution. Astrometry.net does OK with finding the center (tangent point) but not so well at generating a global WCS solution. 

Tried hacking the header of the image that went through astrometry.net, but didn't manage to get Simbad stars to line up in any sensible way. 

So for photometric calibration, let's try something different- namely find a Simbad star that passes right overhead, and find it in the appropriate image. We know the following:

  • fits files are N up E left
  • Image ut011414.0360 was taken at 03:40 UT on Jan 15, and is at RA=102.345 and dec=-28.57. 

cp65239:calib cstubbs$ awk '{if (($2<-25)&&($2>-30)&&($3<8)) print $0}' SimbadBVR.dat > Simbad.zenith.dat

cp65239:calib cstubbs$ wc Simbad.zenith.dat 

      12      60     588 Simbad.zenith.dat

cp65239:calib cstubbs$ cat Simbad.zenith.dat 

096.18283193 -28.78011494    6.99   6.374  6.24 

110.31183 -29.29997          6.75  11.2    9.500

118.45526891 -26.26208578    7.85   6.92   6.48 

118.2697246 -26.4142169      7.45   7.55   7.60 

154.53162512 -28.99199955    5.76   5.53   5.5  

155.98031170 -29.64552930    7.603  6.920  6.534

168.00496442 -26.13667076    7.815  7.037  6.621

173.066850 -29.263294        6.19   5.671  5.49 

194.12554375 -26.46029020    7.35   6.74   6.8  

310.81666184 -29.42391078    7.635  6.960  6.583

228.36944608 -25.30934634    7.168  6.46   6.43 

303.82246344 -27.03297557    6.63   5.723  5.98 

here is this file of bright stars that pass overhead: Simbad.zenith.dat

RA evolves at 15 degrees per hour, and we take 60 images per hour. So an RA of 118.5 is (118.5-102.3)/15=1.08 hours later than frame 380. That's one hour

and 5 minutes later. Using MJD this is 0.0450 days later than image 380, at MJD 56672.161285. So we want 56672.20629. that's image 486

what the heck, try astrometry.net? Nope, failed. 


IRAF MSCTPEAK astrometric fit to ut011214.0650.M.long

IRAF MSCTPEAK data base file: SAO.ut011214.0650.db

SAO Catalog extract covering full image: SAO.ut011214.0650.cat

Conversion  of RA-DEC to Pixel X-Y using the plate solution: SAO.ut011214.0650.xy.cat

Gemini South:

Latitude: -30:14:26 (-30.2406)

Longitude: -70:44:12

UT Date for exposure: 01-13-14 08:39:39

Local Sidereal Time: 11:27:31 HH:MM:SS (170.3792 deg)

Using the above parameters and plate solution the zenith pixel coordinates are x = 1407.48 and y = 923.20.

Modest success with this approach.  Although, have not yet figured out how to get the fit in the database file (.db) into a full high order WCS fits header.

A small section of ut011214.0650 showing the predicted position of identified SAO stars marked with green circles.

Full image with SAO stars marked along with the zenith point based on the plate solution.

 


 

Try a totally different approach. Lens converts alt,az into x,y. Use that as the fundamental representation. 

Generated a couple of Matlab functions, one that goes from x,y to alt,az and the other that goes from alt,az to RA,DEC. use this to test thing out.

ut011414.0306.long.M.fits  2047.889 10.400000 56672.129688      67.4674 63.1341 84.2622 38.4578 1073 is a decent image and photometry file. It was taken on Jan 15 UT at 3.11251 hour night, so 03:06:45. That's a time array of time=[2014 01 15 03 06 45]. 

That didn't work so well.. so let's try going from RA,Dec for Simbad stars to pixel x,y using alt,az as an intermediate coordinate system. Converting Simbad RA Dec to x,y for frame 306 above. 

Nope, plots of stellar positions on image 306 don't make much sense. 

Jan 17 2014. 

One way to debug this positioning code is to get the positions of the sun and the moon right. Using Skycalc for Pachon we get

datesun RAsun Decsun altsun azmoon RA (midnight)moon Dec (midnight)moon altmoon az
Jan 14, noon local19:44:06-21:14:3063.377.006:56:08 jan 14/1518:48:3638.818.7
Jan 15, noon local19:49:24-21:03:3363.276.707:45:42 jan 15/1616:54:0936.632.2
Jan 16, noon local19:53:41-20:52:1163.076.408:34:04 jan 16/1714:18:2032.445.0
Jan 17, noon local19:57:57-20:40:2562.876.109:21:21 jan 17/1811:08:2727.456.3

 

Evidently most of these DSLR fisheye lenses are constructed such that r=alpha*sin(theta/2), with alpha=2f. See http://michel.thoby.free.fr/Canon_8-15mm/8-15mm_review.html which shows


Jan 18 2014. 

Took the daycal images from ut011714 and manually determined the x,y position of the center of the sun. Also used skycalc to determine sky coords of sun for times of observations. Results:

# file                                             DATE-OBS              MJD-OBS    sun x sun y   sunalt  sunaz   sunRA           sunDec          LMST

ut011714.daycal.0001.fits 2014-01-17T12:39:04   56674.527130    219     881     32.6    96.5    299.38667       -20.69278

ut011714.daycal.0002.fits 2014-01-17T12:49:12   56674.534167    258     910     34.8    95.4    299.39417       -20.69139

ut011714.daycal.0003.fits 2014-01-17T12:59:24   56674.541250    305     933     37.0    94.2    299.40167       -20.69000

ut011714.daycal.0004.fits 2014-01-17T13:09:35   56674.548322    343     955     39.2    93.0    299.40917       -20.68861

ut011714.daycal.0005.fits 2014-01-17T13:19:47   56674.555405    383     977     41.4    91.8    299.41667       -20.68722       16.40722

ut011714.daycal.0006.fits 2014-01-17T14:16:47   56674.594988    649     1066    53.7    84.1    299.45875       -20.67944       17.35972

ut011714.daycal.0007.fits 2014-01-17T14:26:55   56674.602025    695     1071    55.8    82.8    299.46583       -20.67806

ut011714.daycal.0008.fits 2014-01-17T14:37:02   56674.609051    748     1089    58.0    80.7    299.47333       -20.67667

ut011714.daycal.0009.fits 2014-01-17T14:47:10   56674.616088    795     1097    60.1    78.8    299.48083       -20.67528

ut011714.daycal.0010.fits 2014-01-17T14:57:18   56674.623125    841     1114    62.3    76.7    299.48833       -20.67389

ut011714.daycal.0011.fits 2014-01-17T15:07:26   56674.630162    900     1095    64.4    74.3    299.49583       -20.67250

ut011714.daycal.0012.fits 2014-01-17T15:54:53   56674.663113    1129    1141    73.8    57.3    299.53042       -20.66611

ut011714.daycal.0014.fits 2014-01-17T16:15:07   56674.677164    1238    1139    77.2    44.0    299.54500       -20.66333

ut011714.daycal.0015.fits 2014-01-17T16:25:15   56674.684201    1285    1140    78.6    34.8    299.55250       -20.66194

ut011714.daycal.0016.fits 2014-01-17T16:35:23   56674.691238    1332    1140    79.6    23.6    299.56000       -20.66056

ut011714.daycal.0017.fits 2014-01-17T16:45:31   56674.698275    1383    1130    80.3    10.5    299.56750       -20.65917

ut011714.daycal.0018.fits 2014-01-17T16:55:38   56674.705301    1444    1139    80.4    356.4   299.57458       -20.65778

 

Let's estimate zenith-pointing pixel. Image ut011314.0360.long.fits was taken at UT 01 14 03:39:57. The LMST then was 06 31 11, which is an RA of 97.79583. Zenith is same as our latitude, namely -30.252090. So, running sky2xy on the 360 wcs header, I get 97.79583 -30.252090 J2000 -> 1407.770  923.766. So let's take the zenith pixel to be (1408,924).  We can use that information to produce the x',y' location of the sun, with x'=x-1408, y'=y-924.

 

That worked well! Best-fit line is r(pixels)=2474.4*sin(zenith/2)+1.6748. How about in the azimuth direction? Take apparent azimuth as being CCW angle from positive y axis to location, centered on (1408,924). That seems OK as well, with apparent azimuth on chip = 0.99765 * azimuth - 4.4343,  where apparent azimuth was computed with app az=(180/pi)*(atan2(-x'.y')), where the order of the argument and the (minus) sign account for the azimuth convention. This is enough to be able to estimate a source's location on the sensor, given its alt,az.

1) compute zenith distance with zenith=90-alt

2) compute pixel radius off zenith, pixrad=2474.4*sind(zenith/2)

3) compute bearing to that location. Pixaz=Azimuth - 4.4 deg

4) compute dx=-1*sindd(Pixaz), dy=cosd(Pixaz)

5) compute x=1408+dx, y=924+dy. 

Test this with AltAz2xy code in matlab: 

AltAz2xy.m

Verified that I can also go the other way, from xy2AltAz:

 

xy2AltAz.m

  • No labels

1 Comment

  1. Unknown User (shaw)

    I'm not surprised that astrometry.net did not work well on these (nearly) all-sky images (see http://astrometry.net/doc/oaq.html). Internally a.n attempts a solution with the TAN-SIP solution, which would not be my first choice of a projection for such a wide area. I suspect you might do better with a ZPN (zenithal polynomial) or ZEA (zenithal equal-area) projection. I think IRAF supports one or the other of these; probably other packages support them too. The solution seems not vastly different from the ideas being explored at the end of this page, except that you can represent the WCS in a FITS header.