The Tractor: Probabilistic astronomical source detection and measurement
Abstract
The Tractor optimizes or samples from models of astronomical objects. The approach is generative: given astronomical sources and a description of the image properties, the code produces pixelspace estimates or predictions of what will be observed in the images. This estimate can be used to produce a likelihood for the observed data given the model: assuming the model space actually includes the truth (it doesn’t, in detail), then if we had the optimal model parameters, the predicted image would differ from the actually observed image only by noise. Given a noise model of the instrument and assuming pixelwise independent noise, the loglikelihood is the negative chisquared difference: (image  model) / noise.
 Publication:

Astrophysics Source Code Library
 Pub Date:
 April 2016
 Bibcode:
 2016ascl.soft04008L
 Keywords:

 Software