Classifying Galaxy Morphology with MatLab
Abel B. Diaz     Michael S. Pate     Ronald Wilhelm
•We are using MatLab to automate galaxy classification
 by extracting the ‘shapes’ of galaxies
•
•The galaxies used were elliptical, irregulars and face-on
 spirals, all large in spatial extent
•
•We present preliminary results for our method as well as
 preliminary results for galaxies of lower resolution
Introduction
Why Classify
vGalaxy Morphology in Rich Clusters: Implications for
   the Formation and evolution of Galaxies, Dressler, A. 1980
vFaint galaxy counts as a function of morphological type
   in a hierarchical merger model, Baugh, C. M. et al. 1996
vColor Separation of Galaxy Types in the Sloan
   Digital Sky Survey Imaging Data,Strateva et al. 2001
Morphology classification provides a direct approach to
probe galaxy populations back to an early stage in their
lifetime, and hence to study many astrophysical
problems such as large-scale structure, galaxy formation
and galaxy evolution.
Current Methods
Most current morphology methods use photometric
Parameters
•concentration index
•mean surface brightness
•bulge-to-disk ratio
•color
•de Vaucouleurs exponential profiles
•
•
A few morphology methods make use of the ‘shape’ of
Galaxies to determine type.
vAutomated Galaxy Morphology:
   A Fourier Approach, Odewahn et al. 2002
vMorphological Classification of Galaxies using
   Computer Vision and Artificial Neural Networks:
   A Computational Scheme, Goderya, S & Lolling, S 2002
Our Method With MatLab
vPreset values: number of profile vectors, background
   star removal and graphical output
vReads in files (jpeg or fits)
vlocates center of image
v
vcreates profile vectors (across the galaxies center) with
   intensity counts
v
vdetermines, from the profile vectors, if galaxy is elliptical,
   irregular, spiral or barred-spiral
What is MatLab?
MATLAB can serve as the platform for all your technical computing needs.
It integrates mathematical computing, visualization, and a powerful technical language. The MATLAB product family includes tools for:
• Test and measurement
• Data analysis and exploration
• Numeric and symbolic computing
• Plotting and advanced visualization
• Signal and image processing
• Algorithm development
• Deployment of MATLAB applications
MATLAB and its companion products are used in a broad range of applications, including signal and image processing, DSP and control design, medical research, finance, and earth and space exploration. For these and many other technical applications, MATLAB gives you a trusted computing environment, fast insights, and a competitive advantage.
More MatLab.
The Image Processing Toolbox enables you to restore noisy or degraded images, enhance images for improved intelligibility, analyze blobs, and extract and analyze image data with image statistics and transforms, as well as develop complete solutions to challenging image processing problems that involve multidimensional data sets.
Test Galaxies and Results
NGC1365 (SBb)
M100 (Sc)
NGC1300 (SBb)
M91 (SBb)
NGC4449 (irregular)
M59 (elliptical)
ngc4348 (edge-on)
M74 (Sb)
NGC2967(Sc)
Resolution Tests
M100 (original)
(less resolved)
(a lot less resolved)
Future Work
vIt was found that if we change some of the program parameters that
   some of the incorrect galaxies would become correct and some that
   were correct become incorrect. We need to automate some of these
   parameters to adjust depending on image characteristics.
v
vWe need to quantify the resolution limits to be able to determine how
   faint we can go.
vFurther morphology subdivision (spiral – Sa, Sb and Sc. Barred-spiral
   SBa, SBb and SBc.) within resolution limit
v
vDetermine morphology (to high accuracy) of galaxies beyond the
    resolution limit by using photometric parameters along with shape
vPossible merging with the procedure described in Morphological
   Classification of Galaxies using Computer Vision and Artificial Neural
   Networks: A Computational Scheme, Goderya, S & Lolling, S 2002
Conclusion
üWe found, using MatLab, it is possible to automate galaxy
   morphology by ‘shape’, without photometric parameters. These
   particular results were about 80% accurate and can be greatly
   improved by changing the program.  
ü
üThis method is not limited to SDSS data, any survey or data will
   work
ü
üThis method does have limits from a resolution stand point.
ü
üWe can extend this limit by using photometric parameters in
   conjunction with ‘shape’.
Contact Info
Abel B. Diaz: abel.diaz@ttu.edu
Michael S. Pate: michael.s.pate@ttu.edu
Ron Wilhelm: ron.wilhelm@ttu.edu
http://www.mathworks.com/
If no arms are found it checks average number of counts in profile vector above ~80%. If average is > 2 then it is labeled an irregular galaxy, if average is < 2 then it is labeled an elliptical galaxy.
If armed structure are found (traced in red) using the profile vectors the galaxy is labeled a spiral. If the ratio of the major and minor axis is > 2 then the galaxy is labeled a barred-spiral.
 
Ir
Ir
sextenA
 
spiral (barred)
SBb
ngc4579
possibly a threshold limit in the program is not small enough for this diffuse SB0
spiral (non-barred)
SB0
ngc4477
 
Ir
Ir
ngc4449
 
spiral (non-barred)
Sb
ngc4348
 
spiral (non-barred)
Sb
ngc4030
 
spiral (barred)
SBc
ngc309
 
spiral (non-barred)
Sc
ngc2967
 
spiral (barred)
SBb
ngc165
 
spiral (barred)
SBb
ngc1365
 
spiral (barred)
SBb
ngc1300
there is a small galaxy near by that was used in determining if it was an irregular
Ir
SBb
ngc1097
 
spiral (barred)
SBb
ngc1090
 
spiral (non-barred)
Sa
m96
the ring structure could have been used for bar calculation
spiral (non-barred)
SBb
m95
the spiral arms back track some and could have been used for bar calculation
spiral (non-barred)
SBb
m91
 
E
E
m87
This is a questionable barred. The bar srtucture is mostly gas not stars.
spiral (non-barred)
SBc
m83
 
Ir
Ir
m82
 
spiral (non-barred)
Sc
m74
 
E
E
m59
possibly couldn't separate bulge from the arms
spiral (barred)
Sc
m51
 
spiral (non-barred)
Sc
m33
 
spiral (non-barred)
Sc
m100
comments of what caused the incorrect type
type (program)
type
name
 
 
from www.seds.org
 
Summary of Results
We found that if changes are made to the program parameters that some of the incorrect
galaxies would become correct and some that were correct become incorrect.
We found that the program could still find spiral structure with low resolved images.
 (resolution to low/no arms found)