Star/Galaxy Separation Revisited : Into the Zone of Avoidance

Abstract

The problem of automated separation of stars and galaxies on photographic plates is revisited with two goals in mind : First, to separate galaxies from everything else (as opposed to most previous work, in which galaxies were lumped together with all other non-stellar images). And second, to search optically for galaxies at low Galactic latitudes (an area that has been largely avoided in the past). This paper demonstrates how an artificial neural network can be trained to achieve both goals on Schmidt plates of the Digitised Sky Survey. Here I present the method while its application to large numbers of plates is deferred to a later paper. Analysis is also provided of the way in which the network operates and the results are used to counter claims that it is a complicated and incomprehensible tool.