search

PaulHancock / Aegean

The Aegean source finding program and associated tools
http://aegeantools.rtfd.io/
Other
47 stars 14 forks source link

Insight or guidance to improve bad results #113

Open caseyjlaw opened 4 years ago

caseyjlaw commented 4 years ago

I'm applying aegean for the first time to all-sky images from the OVRO-LWA at 60 MHz. I am visualizing in saoimage and overlaying the output region files. I'm seeing some sources fit well, some not so well. The performance seems to have little to do with the source brightness or proximity to neighboring sources. I'm at a loss to explain some of the results, so I thought I'd raise an issue to ask for help in understanding what drives bad behavior in the source detection/fitting. I'm not sure if this is a request for documentation (which is already great) or new features, but I thought I'd throw this out there. An example image near the center of the field is at http://www.aoc.nrao.edu/~claw/aegean_example.png. Two common patterns in the image shown here and elsewhere in this image:

Any idea how I can get insight to why aegean is finding sources this way?

PaulHancock commented 4 years ago

Those examples are rather disturbing. The field doesn't look overly complicated, and the sources are in what I consider the 'easy to fit' category (isolated unresolved sources). I would expect Aegean to be excellent in this situation.

Are you able to provide:

This should help to pick up any obvious issues.

On a side note: I recently made some changes to the way that Aegean does the psf calculation, so the most recent version (v.2.2.0) may give you better results than any previous version (v2.1.0 or earlier).

caseyjlaw commented 4 years ago

1) Typically I'm running like this: > aegean int00001-image_medfilt.fits --out int00001-image_medfilt.csv --table int00001-image_medfilt.reg --seedclip 7 --floodclip 5 --cores 1

I've also tried making using all cores and my own background/rms images, but the quality of results do not change radically.

2) Versions:

aegean --versions
42379:aegean:INFO This is Aegean 2.1.0-(2019-10-06)
42379:aegean:INFO AegeanTools 2.1.0 from /Users/claw/anaconda3/envs/py36/lib/python3.6/site-packages/AegeanTools/__init__.py
42379:aegean:INFO Numpy 1.18.1 from /Users/claw/anaconda3/envs/py36/lib/python3.6/site-packages/numpy/__init__.py 
42379:aegean:INFO Scipy 1.4.1 from /Users/claw/anaconda3/envs/py36/lib/python3.6/site-packages/scipy/__init__.py
42379:aegean:INFO AstroPy 4.1.dev27307 from /Users/claw/anaconda3/envs/py36/lib/python3.6/site-packages/astropy-4.1.dev27307-py3.6-macosx-10.7-x86_64.egg/astropy/__init__.py
42379:aegean:INFO LMFit 0.9.14 from /Users/claw/anaconda3/envs/py36/lib/python3.6/site-packages/lmfit/__init__.py

3) fits header

fitshdr int00001-image_medfilt.fits 
================================================================================
FITS header number 1 at block number 1.
--------------------------------------------------------------------------------
SIMPLE  =                    T / conforms to FITS standard
BITPIX  =                  -32 / array data type
NAXIS   =                    4 / number of array dimensions
NAXIS1  =                 4096
NAXIS2  =                 4096
NAXIS3  =                    1
NAXIS4  =                    1
COMMENT   FITS (Flexible Image Transport System) format is defined in 'Astronomy
COMMENT   and Astrophysics', volume 376, page 359; bibcode: 2001A&A...376..359H
BUNIT   = 'JY/BEAM '           / Units are in Jansky per beam
BMAJ    =      0.4794409279927
BMIN    =    0.224118040635636
BPA     =     49.2725776728388
EQUINOX =                2000. / J2000
BTYPE   = 'Intensity'
ORIGIN  = 'WSClean '           / W-stacking imager written by Andre Offringa
CTYPE1  = 'RA---SIN'           / Right ascension angle cosine
CRPIX1  =                2049.
CRVAL1  =             8.758917
CDELT1  =             -0.03125
CUNIT1  = 'deg     '
CTYPE2  = 'DEC--SIN'           / Declination angle cosine
CRPIX2  =                2049.
CRVAL2  =             36.95993
CDELT2  =              0.03125
CUNIT2  = 'deg     '
CTYPE3  = 'FREQ    '           / Central frequency
CRPIX3  =                   1.
CRVAL3  =            56148000.
CDELT3  =            57552000.
CUNIT3  = 'Hz      '
CTYPE4  = 'STOKES  '
CRPIX4  =                   1.
CRVAL4  =                   1.
CDELT4  =                   1.
CUNIT4  = '        '
SPECSYS = 'TOPOCENT'
DATE-OBS= '2017-01-12T01:02:10.6'
WSCDATAC= 'CORRECTED_DATA'
WSCWEIGH= 'Briggs''(0)'
WSCFIELD=                   0.
WSCGAIN =                  0.1
WSCGKRNL=                   7.
WSCIMGWG=     20209356.6913558
WSCMAJOR=                  12.
WSCMGAIN=                 0.85
WSCMINOR=               50000.
WSCNEGCM=                   1.
WSCNEGST=                   0.
WSCNITER=               50000.
WSCNWLAY=                  32.
WSCTHRES=                   0.
END
--------------------------------------------------------------------------------
Data section number 1 beginning at block number 3.
Skipped 23302 blocks of data of size 2880 bytes (67109760 bytes)..   
================================================================================
End-of-file after 1 HDU in 23304 x 2880-byte blocks (67115520 bytes).
--------------------------------------------------------------------------------
caseyjlaw commented 4 years ago

I tried running version 2.2.0, but ran into a different issue (runs forever and memory use blows up at "Begin fitting"). I can submit an issue on that if you like.