ASTAP plate solver use in SGP


#1

ASTAP is a new free astro program having a native internal plate solver. It is command line compatible with PlateSolve2.exe 2.29 and can work with SGP as a PlateSolve2.exe substitute by simple replacing the PlateSolve2.exe executable and adding the star database in the same directory… The ASTAP plate solver is an own idevelopment, not depending on an external programs.

It has the following advances:

  • The lastest version 0.9.114 can do blind solving in minutes.
  • Not dependent on Cygwin. Installation is simple.
  • Doesn’t require an accurate image size in degrees. Only roughly the image height in degrees. Much more tolerant then PlateSolve2.
  • Command line compatible with PlateSolve2.exe but ASTAP has an own command line {preferred}

Please try it and if you think it is beneficial, I would suggest to add ASTAP as an option to the SGP Plate Solve menu.

Home page:
http://www.hnsky.org/astap.htm

Command line:
http://www.hnsky.org/astap.htm#command_line

PlateSolve substitute:
http://www.hnsky.org/astap.htm#platesolve2

Astap can do more but that is not of interest here.

I’m the programmer of both ASTAP and HNSKY planetarium program. There is no commercial interest. The programs are free to use.
Han
The Netherlands.
www.hnsky.org


#2

Hi Han,

This sounds awesome! Can you please provide comparison between ASTAP, ANSVR and PlateSolve 2? Why should we choose ASTAP instead of anything else?

I like that your program is platform independent almost. You even provide source code which is awesome. What do you think about making the code available on a platform like github instead of dropbox? Makes source control easier and people can contribute if they’d like.

Hope to hear your thoughts.


#3

Hello Zwik,

The program is also available at sourceforge:
https://sourceforge.net/projects/astap-program/

Probably the best download link for high demands.

I’m a little reluctant to make a calculated comparison. PlateSolve2.exe and Astrometry.net based programs are all robust solvers. I’m confident now this also applies for ASTAP. Performance of Astrometry.net is very dependent on the settings. Each solver has it’s strengths. But generic I can say the following:

For small offsets, ASTAP solver will beat Astrometry.net in speed.

For large offsets, ASTAP will beat PlateSolve2. A test a few day ago with a 50 seconds exposure starting with a 180 degrees offset, so a full sky scan so through both Northern and Southern hemisphere took 380 seconds. So a blind solve starting with an offset of 90 degrees will be found in 190 seconds. So in practice an offset of 30 degrees will be found quickly. PlateSolve2 can’t do that. This by a brute force method, not by previously prepared index files as for Astrometry.net.

For problematic images, probably Astrometry.net is the best. For short exposures of a few seconds taken in twilight, ASTAP and PlateSolve2 have a a very similar performance. The star detection routine in ASTAP is robust.

Astrometry.net and ASTAP don’t need a correct image scale. An image scale offset of 50% can be acceptable for ASTAP. PlateSolve2 is very critical on the image scale.

Maybe not of interest for SGP, but ASTAP native executables are available for MSWindows and Linux. PlateSolve2 and Astrometry,net require an virtual environment.

Both Astrometry.net and ASTAP provide deepsky annotation. The deepsky database of ASTAP is better (based on HNSKY database)

Rather then judging my arguments, I would suggest to try it. Execute ASTAP and load a FITS image in the viewer and hit the solve button. Better hit the Σ button(ctrl+A), tab ALIGNMENT, button “Find astrometric solution current image” At the bottom have a look to the log. In this tab you can set a few things like binning for images larger then 3000 pixels wide but most likely no more settings changes are required. If you load an JPEG image, provide the expected center position in the viewer and the image height in degrees in the tab ALIGNMENT.

I you happy with the performance, rename astap.exe to PlateSolve2.exe and place it in the users directory to substitute PlateSolve2. Add the 290 files of the G17 star catalog in the same directory. Activity will be show by a small tray icon only. Confidence will be reported as 999 indicating ASTAP is reporting.

Note of warning, If you imaging at more then 3000 pixels wide, first set binning at x2 or x3 in the tab Alignment and leave the program with FILE, EXIT to save this.

ASTAP is still in active development and still beta, but soon will be at version 1.0. First solver version was released maybe in April 2018. Improvements are still included.

Note that:
The internal astrometric solver works best with raw unstretched and sharp images of sufficient resolution where stars can be very faint. Heavily stretched, very long exposures or photo shopped images are problematic. It requires minimum about 30 stars in the image. Images containing of a few hundred stars are ideal. For star rich images it reduce the detection limit to limit the number of stars. This will only work for unstretched images.

This became a long piece. Hope this helps.

Han

.


#4

well this is not entirely true, at least for astrometry.net. it compiles on cygwin and under macosx and linux just fine. no virtual machine is required, unless you are counting cygwin as a virtual environment.


#5

I did.:slight_smile: For Windows to get Astrometry.net running you will need Cygwin or the Win10 64 bit Linux sub-system. Both Astrotortilla and ANSVR use the same old Astronomy.net v0.38 somebody managed to compile long time ago. In the Win64 special Linux sub-system it is possible to install a later version of Astrometry.net See my guideline:

http://www.hnsky.org/linux_subsyst.htmhttp://www.hnsky.org/linux_subsyst.htm

However this last setup is not accessible by SGP since it doesn’t use the Astrometry.net command line but communicates via the ANSVR server.

For Astrometry.net they choose a mix of C language programs and Python code making porting a little difficult.

ASTAP is purely written in Object Pascal and use Lazarus FPC to compile making compilation to Windows, Linux and OSX easy. However Object Pascal is not so main stream.

I think in the long term, it would be wise to add in SGP an option to use the command line of Astrometry.net so it would be able to access a local up to date Astrometry.net in a Win10 64 bit Linux sub-system. But they should also try ASTAP :slight_smile:

I maybe have typed a too long story, but i have busy in this topic for some time. Hopes this helps.

Han


#6

I just like to add how this all started. Originally ASTAP started as a stacking program (and still is) and I used Platesolve2 to get an astrometrical solutions to stack images. The accuracy of PlateSolve2 was not satisfactory for stacking, so I added access to Astrometry.net. That worked fine but slow. Then I developed a star matching routine for stacking which worked much faster and the need for Astrometry.net was limited for Mosaic building where star matching no longer works.

Still impressed by plate solving capabilities of Astrometry.net and PlateSolve2, and having star matching for stacking working, I decided in the spring of 2018 to try to develop my own solving routines.


#7

Version 0.9.117 just released is about twice as fast.

ASTAP blind solver performance, solving a 50 seconds exposed monchrome image of M24, 2328x1760 pixels covering a field of 1.75 x 1.32° starting 90 degrees more north at position 18:17, +72d 00

Max stars set 	Tolerance set	Astrometric solving  time
500		0.005		86 sec.
250		0.005		40 sec
150		0.007		31 sec
100		0.007		21 sec

Reducing to the brightest stars only will result in increasing solving speed but increased risk of missing the solution.Above result show that blind solving with brute force can give excellent results.


#8

Unfortunately ASTAP version 0.9.117 has a bug in reading the star database and has been replaced by 0.9.118.

Speed for blind solving has unfortunately been reduced by correct reading the star database. Images with small offsets are still solved within seconds:

ASTAP blind solver performance, solving a 50 seconds exposed monchrome image of M24, 2328x1760 pixels covering a field of 1.75 x 1.32° starting 90 degrees more north at position 18:17, +72d 00

Max stars set 	Tolerance set	Astrometric solving  time
500		0.005		147 sec.
250		0.005		103 sec
150		0.005		92 sec
100		0.005		70 sec

Reducing to the brightest stars only will result in increased solving speed but an increased risk of missing the solution. Above result show that blind solving with brute force can give excellent results.

I also rechecked in simulation SGP using ASTAP as PlateSolve2 substitute and could not find any problem. You will only notice ASTAP by the confidence reporting 999. I’m not an SGP user but in my setup the solver has not failed me in the past two months.

Feedback and questions and future updates, I would like to refer to the ASTAP forum rather then this SGP forum. I don’t want to hijack this forum.

Han


#9

Below a solving speed comparison between PlateSolve2 and ASTAP 0.9.127:

Offset_____PlateSolve2____ASTAP 0.9.127
0°________1.5 sec___________1.4 sec
2°________ 15 sec___________1.5 sec
5°_______122 sec____________2 sec
10°______535 sec____________4 sec
30°_______________________18 sec
60°_______________________97 sec

(mono image of 10 seconds exposure containing about 200 stars, dimensions 2328x1760):

IMPORTANT, previous version of ASTAP had bug which only occured for DSLR cameras. ASTAP looked only to the initial start position in the FITS header of the input file and not to the start position given by SGP in the PlateSolve2 command line. This resulted in long solving times. This has been fixed in ASTAP version 0.9.127 just released.

Latest version also produces a log file in the documents folder as follows:

Solved,C:\Users\H\AppData\Local\SequenceGenerator\Temp\psXSolve_11.fit, Solved in 1.4 sec., Offset was 0.090 deg.
Solved,C:\Users\H\AppData\Local\SequenceGenerator\Temp\psXSolve_12.fit, Solved in 1.5 sec., Offset was 2.001 deg.
Solved,C:\Users\H\AppData\Local\SequenceGenerator\Temp\psXSolve_13.fit, Solved in 4.5 sec., Offset was 4.922 deg.
Solved,C:\Users\H\AppData\Local\SequenceGenerator\Temp\psXSolve_14.fit, Solved in 3.9 sec., Offset was 10.011 deg.
Solved,C:\Users\H\AppData\Local\SequenceGenerator\Temp\psXSolve_15.fit, Solved in 17.3 sec., Offset was 30.011 deg.

This log file grows till it is removed. The file names are from SGP.


#10

This sounds really promising and I will experiment with it soon when time allows. What do you think about porting to another language that is a bit more mainstream? I looked into your code on SourceForge but it’s very hard to read, variables that have no meaningful name etc. make it really hard to read.


www.mainsequencesoftware.com