You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 9 Next »

Describe the process of "detection", or finding objects in individual PanSTARRS exposures, and the quantities associated with a detection.  Detections are combined into "objects" by spatial matching across different exposures and filters. 

 

Contents

Definition - what is a "detection"?

In PanSTARRS lingo, a detection is a source found in a specific exposure.  Detections are matched across exposures to define "objects".

Detections are identified through a standard peak-finding algorithm.  The image is convolved with an approximation of the PSF and then divided by a smoothed version of the variance image to define the significance image.  Peaks are defined as locations where the significance image exceeds a target threshold, representing the square of the desired signal-to-noise ratio.  Peaks are then ordered in decreasing significance, and peaks are retained only if a significant valley separates them from brighter nearby peaks.

 

The process of identifying detections is complex and involves multiple steps.

  • Smooth Image with PSF (or guess in first pass)
  • Smooth Variance with PSF2
  • (Actually, I cheat & use a 1D Gaussian with FWHM matching PSF)
    • much faster, marginal difference
    • if the difference matters, image is quite poor quality
  • Significance image by dividing image2 / variance
  • Find all peaks above target S/N (squared)
  • Footprint analysis
    • generate isophotal footprint outlines (N sigma above sky)
    • assign peaks to their containing footprints
    • cull insignificant peaks
      • cull in descending brightness order
      • valid peak must be separated from a brighter peak by a significant valley
      • recent fix : on second pass, cull on the unsubtracted image

Some information from Tonry 2012a (white dwarfs from MDF data) to be placed in this page tree. 

1) Nightly stacks.  Images obtained from a signle night in each band are typically obtained with small changes in boresight and at a variety of rotator angles.  Eight images were obtained per filter per night, using a variance-weighted combination of individual frames and outlier rejection.  Nightly stacked images are considered to be that night's image in the appropirate band.  (Flux is adjusted at the level of a Pan-STARRS skycell, about 20 arcmin on the sky.)  If this process is applicable to all 3PI data, the concept of "image" has to be clarified.

2) Stacking.  The process Tonry describes involves weighting images from different nights by their inverse variance times the inverse PSF area, nearly optimal for point-source detection.  No convolution is used, thus the effective PSF for a point source will have a sharp core due to images obtained in better seeing, and a skirt due to poorer seeing.  He claims that deconvolution of the skirt would be possible, but it was not attempted.

3) Photometric calibration.  Tonry lists a sequence of steps used to bring different images to a common zero point. 

a) obtain positions and fluxes for all stars in each nightly stack using DOPHOT

b) obtain photometric and astrometric offsets for each skycell by pairwise comparing data for all sources. (Rejection?)  Instrumental magnitudes are defined for a 6" box.

c) stack nightly stacks as described above

d) Obtain instrumental magnitudes for the stats in a stack-stack

e) obtain a zero point for the stack-stack by comparison with SDSS-converted-to-PanSTARRS1, 2MASS+stellar locus correction,

pairwise

 

PSF Photometry

 

PSF fitting

 

Photometric zero points

 

Aperture Photometry

 

Kron Photometry

image smoothing, sky level, measuring moments, aperture size, iterations

 

Forced Photometry

 

  • No labels