Integration of All Sky Plate Solver


#1

I would like to suggest to integrate All Sky Plate Solver by Giovanni Benintende http://www.astrogb.com into SGPRO. It is a very simple to handle Blind Solver
based on astrometry.net (local). It is really speedy and very reliable. Could be a
fantastic alternative to Elbrus which is a bit cumbersome in handling and somewhat outdated.
I hassled with a lot of plate solving solutions in the past, but this is by far the best i have seen so far.

Thanks for your good work.

Robert


#2

@winwimp Thanks for the suggestion. That said it is not clear to me what advantage this would add over @Andy 's ANSVR. Maybe I’m being dense. Is there one?


#3

Adding a link to ansvr in case @winwimp was not aware of it.
Andy


#4

Hi Ken,

yes, at least for my environment ASPS is faster and finds a solution in
almost all
cases without tweaking. Ansvr often fails where ASPS solves the plate.
Since elbrus
seems to be no longer supported, ASPS could be an alternative to it.
Go and try it, installation is a snap when you have already installed the
indexes.
The software works standalone with an interface to Maxim or Ascom-camera.
Integration doesn’t seem to be too difficult since Ivo (APT) has already
dunnit.

@andy: Maybe something is faulty in my ansvr installation. I work with the
default
parameters in the web interface. What i changed is that i kicked off
astrotortilla and
upgraded to Win10. I have some fits files where ansvr never finds a
solution (even with
RA and DEC values in the fits header) whereas ASPS crumbles it in 25 sec.

Thanks
Robert


#5

Hi Robert,

ASPS is using the exact same plate solver as ansvr, astrometry.net 0.38. Any difference in plate solving results has to be a result of passing different options to astrometry.net or of not having the right set of index files. The ASPS log does not show the exact set of options that ASPS is passing to solve-field, though we can deduce it is using a downsample factor of 4.

I will contact Giovanni and see what settings he is using. In the meantime, could you send me a link to one of your fits files that solves faster using the ASPS settings?

Andy


#6

Ok, I found out what command-line options ASPS is using.

If you configure ansvr like this, you should get the same results as ASPS

You should also confirm that:

  • you have the correct focal length and pixel scale specified in SGP
  • you are using the same set of index files in ansvr that you were using in ASPS

Please let me know if you still see any difference and we’ll track it down.

Andy


#7

Hi Andy,

I changed the settings according to your proposal. Still 140s to go for
ansvr
whereas ASPS finds the coordinates in 24s.
Indexes installed are 4205 - 4215 but these are used by both apps.
Maybe its time to rip off the whole bunch of apps from cygwin to ansvr
and give it a try with a clean start.
If you need some data for analyzing let me know.
Here is the link to the fits

C u on the bitstream
Robert


#8

Hi,
The image scale is missing.
If you supply the image scale 2.31
Ansvr solves in 12 seconds
PS2 solves in 4 seconds
PS2 will even solve if you use an image scale of 2
Nigel


#9

Hi Nigel,

you are right, with the correct image scale it solves in a few seconds.
I wonder how ASPS manages it. With wrong camera pixel size
and right focal length it solves a plate in 26 seconds, with a rough
image scale ansvr has got a loooong way.

Robert


#10

Thanks, Nigel.

I agree, the image scale is the critical element. ASPS requires you to explicitly provide focal length and pixel size. With SGP/ansvr, you have the option to leave the scale unspecified in which case astrometry.net will search all image scales for which you have indexes and that can be very slow.

I ran 7 consecutive solves of the IC805 image with the correct image scale specified and using the objs = 100 and downsample = 4 settings (so ansvr is using the same settings as ASPS). Here are the timings I got, in seconds:

ansvr: 12, 11, 9, 5, 5, 5, 5
ASPS: 9, 7, 7, 7, 7, 7, 7

both solvers get faster after a few solves because the index files get cached in.memory after the first solve or two.

ASPS and ansvr are both using the identical same underlying plate solving engine, astrometry.net 0.38. Any difference in timings is due to the solver running with different options (including scale hint) and/or index files.

Andy


#11

Hi Andy,

if the image scale is the critical element, then this value should be extracted (calculated) from the fits header or the Equipment profile. Maxim D/L stores these values in the fits header so that Pinpoint can
calculate the image scale. Also the wcs coordinates are stored there after an successfull plate solve.
Some scripts in Pixinsight for example can use these values.

Robert


#12

Good question, but Ken or Jared will need to answer it since it is an SGP question.

Andy


#13

Sorry for being dense but I’m a little unsure what the question is here, so I’ll just take a guess at what is being asked.

SGP stores all data necessary to plate solve an image with an external application in the FITs header. PROVIDED that the information is accessible to us (ie telescope is connected and arcsec/pixel info has been set on the camera or provided by a platesolve within SGP).

So if you open up the FITS header from an image that has been taken with SGP you’ll see the wcs coords, sky angle, scaling, as well as a ton of other information.

Thanks,
Jared


#14

Hi Jared,

right guess. If an image has been taken with SGP
after a successfull plate solve this could be written to the header as well
i.e.
HISTORY File was processed at 2016-03-02T17:13:08
EXPTIME = 6.00000000000E+002 / [sec] Duration of exposure

DATE = ‘10/10/15’ / [old format] UTC date of exposure start

TIME-OBS = ‘20:48:08’ / [old format] UTC time of exposure start

UT = ‘20:48:08’ / [old format] UTC time of exposure start

TIMESYS = 'UTC ’ / Default time system

RADECSYS = 'FK5 ’ / Equatorial coordinate system

CLRBAND = 'R ’ / [J-C std] Std. color band of image or
C=Color
FWHM = 6.00000000000E+002 / [pixels] Mean Full-Width-Half-Max of image
star
ZMAG = 1.57031740690E+001 / Mag zero point for 1 sec exposure

EQUINOX = 2000.0 / Equatorial coordinates are J2000
EPOCH = 2000.0 / (incorrect but needed by old programs)
PA = 2.65391611193E+002 / [deg, 0-360 CCW] Position angle of plate

CDELT1 = 6.42508175242E-004 / [deg/pixel] X-axis plate scale

CROTA1 = -8.53916111928E+001 / [deg] Roll angle wrt X-axis

CDELT2 = 6.42471349447E-004 / [deg/pixel] Y-Axis Plate scale

CROTA2 = -8.53916111928E+001 / [deg] Roll angle wrt Y-axis

CD1_1 = 5.16222320635E-005 / Change in RA—TAN along X-Axis

CD1_2 = 6.40394320310E-004 / Change in RA—TAN along Y-Axis

CD2_1 = -6.40431027051E-004 / Change in DEC–TAN along X-Axis

CD2_2 = 5.16192733001E-005 / Change in DEC–TAN along Y-Axis

TR1_0 = 1.66800022395E+003 / [private] X-axis distortion coefficients

TR1_1 = 3.33599995410E+003

TR1_2 = 5.59855972870E-001

TR1_3 = 2.45189695602E+000

TR1_4 = -7.48476807362E-001

TR1_5 = 1.28336963039E+000

TR1_6 = -1.56693503659E+001

TR1_7 = -1.42415559999E+000

TR1_8 = -1.09928192796E+001

TR1_9 = 2.41483387132E-001

TR1_10 = -3.46008184697E+000

TR1_11 = -2.29046736602E+000

TR1_12 = 3.02630780477E+000

TR1_13 = 2.07902542495E+000

TR1_14 = -4.04578054559E+000

TR2_0 = 1.24799999923E+003 / [private] Y-axis distortion coefficients

TR2_1 = -2.15887070091E-003

TR2_2 = 2.49599999856E+003

TR2_3 = 1.98828330133E+000

TR2_4 = -5.53282816164E-001

TR2_5 = 2.67191081421E-002

TR2_6 = -2.13377648093E+000

TR2_7 = -1.37491134884E+001

TR2_8 = 3.01131415852E-001

TR2_9 = -7.57817208613E+000

TR2_10 = -8.69994128133E+000

TR2_11 = 5.76282577602E+000

TR2_12 = -4.63022427728E+000

TR2_13 = 6.39416054747E-001

TR2_14 = -2.07588885142E+000

HISTORY WCS added at 2016-03-02T17:13:09
HISTORY Matched 392 stars from the Gray GSC-ACT Catalog

HISTORY Average residual was 0.47 arc-seconds

If the image was taken with another software the image scale (if not present in the header)
could be calculated from the given values and written to the solver window.

Robert


#15

Just wanted to say I set the downsampling to 4 and objects to 100 and decreased my already fast plate solving by half! Thanks! I’ll keep a note about these settings in case of any issues though.


#16

Can someone tell how to add the image scale?

I don’t see it in the HTML dialog that ansvr uses.

Max


#17

It’s on the camera tab in SGP. Setting it in ANSVR would mean having to redo plate solver settings for each different scope/camera combo.

Jared


#18

Eh, font scaling win (?).

Jared


#19

Ok, I had this filled out.
Blind solver is still very slow but at least I have this setting covered.

Max


#20

[quote=“maxm, post:19, topic:3144”]
Blind solver is still very slow but at least I have this setting covered.[/quote]

Are you meaning that “Solve and Sync” is faster than the blind option when using ansvr as your only solver? Or just that plate solving is slow in general with ansvr?

Jared


www.mainsequencesoftware.com