Skip to content

GSoC 2016 Application Zé: Implement PSF photometry for fitting several overlapping objects at once

Jump to bottom Edit New page
Zè Vinícius edited this page Apr 2, 2016 · 2 revisions

Student Information

Project Proposal Information

Proposal Title

PSF photometry for fitting multiple overlapping objects simultaneously

Proposal Abstract

Aperture photometry assumes that the background varies in a linear fashion in the aperture’s vicinity. However, in a dense star cluster the background is usually nonlinear. Therefore, one may use point spread function (PSF) photometry in order to meaningfully measure the brightnesses of the sources. In the latter approach, a single model (usually Gaussian shaped) is fitted to each object allowing one to determine, with subpixel precision, their position (x, y center), amplitude, and width. The uncertainties associated with each of these parameters may be determined as well.

However, this becomes a non straightforward task when considering fitting multiple overlapping objects. To do so, one can not “just fit a model with hundreds parameters”. In fact, there exist several problems with this “brute force” approach, and the most critical one might be that the parameter space will have many dimensions (as many as the number of parameters, precisely), which almost certainly will make optimization algorithms to diverge or to get stuck in a local minima.

Hence, my primary task is to work on developing an algorithm to localize, fit, and perform photometry of several overlapping objects (e.g. a dense star cluster, globular clusters, etc) simultaneously.

Proposal Timeline

For clarification, the community bonding period starts on April 22, the first week of coding starts on May 23, and the last week of coding starts on August 23. I'm not committed with any other activity during summer, however I will be attending a conference in the United States from June 26 to July 1. Writing tests and documentation are implicitly included on the proposed timeline.

Community Bonding Period

  • Meeting with mentors via Skype. Meeting other photutils and astropy core developers, and the develop community in general, in order to gather suggestions to build a concrete plan of action before actually starting writing code.
  • Perform a thorough bibliographical research to see how astronomy softwares have been tackling the problem of fitting multiple overlapping sources. For now, the classic DAOPHOT fitting routines for crowded fields will certainly be implemented.
  • Discuss design API and interface.

Week 1

  • Get acquainted with PSF photometry features of photutils.

Weeks 2-4

  • Implement and validate a similar algorithm as presented in the classic DAOPHOT fitting routines for crowded fields [1].

Weeks 5

  • Prepare for midterm evaluation. Start to implement and validate at least two recent published PSF photometry algorithms for crowded fields.

Week 6

  • I will be attending a conference from June 28 to June 30. I will work on Sundays to make up for this.

Weeks 7-8

  • Implement and validate at least two recent published PSF photometry algorithms for crowded fields (continuation).

Week 9

  • Implement benchmarking and PSF uncertainties functions.
  • Starting cythonize or parallelize any code, if necessary.

Week 10

  • Compare the performance of the implemented PSF photometry algorithms with respect to computational time, uncertainties, number of correctly detected sources, etc, using artificial and real data (e.g., Spitzer).
  • Continue to cythonize or parallelize any code, if necessary.

Week 11-12

  • Write notebooks with examples/tutorials demonstrating the use of the developed fitting procedures.

Weeks 13-14

  • Improve documentation, improve tests, debug, and final evaluation.

If time allows

  • Research and implement other source detection algorithms, PSF models, PRF photometry, optimization algorithms, etc...
  • Implement camera-specific (CCD, EMCCD, and sCMOS) maximum likelihood fitters, including uncertainties from the Fisher Information Matrix. See [2], [3]. This makes a huge difference in extreme scenarios (low counts or high noise)
  • Defocused PSF Photometry

A Posteriori GSoC

Keep sending PRs to Astropy :)

Link to Patches

References

[1] P. B. Stetson. DAOPHOT - A computer program for crowded-field stellar photometry. Publications of the Astronomical Society of the Pacific, 99:191–222, March 1987.

[2] J. Skottfelt, D. M. Bramich, R. Figuera Jaimes, U. G. Jørgensen, N. Kains, K. B. W. Harpsøe, C. Liebig, M. T. Penny, K. A. Alsubai, J. M. Andersen, V. Bozza, P. Browne, S. Calchi Novati, Y. Damerdji, C. Diehl, M. Dominik, A. Elyiv, E. Giannini, F. Hessman, T. C. Hinse, M. Hun- dertmark, D. Juncher, E. Kerins, H. Korhonen, L. Mancini, R. Martin, M. Rabus, S. Rahvar, G. Scarpetta, J. Southworth, C. Snodgrass, R. A. Street, J. Surdej, J. Tregloan-Reed, C. Vilela, and A. Williams. EMCCD photometry reveals two new variable stars in the crowded central region of the globular cluster NGC 6981. Astronomy & Astrophysics, 553:A111, May 2013.

[3] P. Qiu, Y.-N. Mao, X.-M. Lu, E. Xiang, and X.-J. Jiang. Evaluation of a scientific CMOS camera for astronomical observations. Research in Astronomy and Astrophysics, 13:615–628, May 2013.

Add a custom footer
Add a custom sidebar
Clone this wiki locally