Automated Source Classification using a Kohonen Network

Abstract

We report progress in the development of automatic star/galaxy classifier for processing images generated by large galaxy surveys like APM. Our classification method is based on neural networks using the Kohonen Self-Organizing Map approach. Our method is novel, since it does not need supervised learning, i.e. human factor, in training. The analysis presented here concentrates on separating point sources (stars) from extended ones. Using simple numerical experiments we compare our method of image classification to the more traditional (PSF-fitting) approach of DAOFIND.