Automated classification techniques of galaxies using artificial neural networks based classifiers

Abstract

Processing and classifying galaxy information is one of the most important challenges and intensive research area for astronomers. In this paper; analyzing and classifying photographic images of galaxies are presented, with interesting scientific findings gleaned from the processed photographic data. In addition, the performance of ten artificial neural networks (ANNs) based classifiers was evaluated, based on a selected set of features. They are a combination of a set of morphic features; derived from image analysis and principal component analysis (PCA) features. These features are combined and arranged to constitute five groups of features. The results showed that; the support vector machine (SVM) based classifier provides the best results; about 99.529% for a feature set composed of the nine morphic features and 24 principal components; occupying 85% of the original data. The dataset was ten cases of NGC category taken from standardized catalog from Zsolt Frei website.

Galaxy Image Classification using Non-Negative Matrix Factorization

“…Galaxies show a vast range of forms, Astronomers for many decades have been trying to find the underlying order in the diverse properties of galaxies [6]. They have found that galaxies can be categorized based on their physical appearances.…”

“…They have found that galaxies can be categorized based on their physical appearances. One of the most common classification schemes is the system devised by Sir Edwin Hubble in 1936, It is often known as the "Hubble tuning-fork", Hubble"s scheme divides galaxies into three classes based on their visual appearance: (i) Elliptical; (ii)Spiral; (iii) Barred spiral; and (iv)Irregular [6]. As shown in figure (1).…”

“…These barred spirals are given the symbol "S.B.".  Lenticular galaxies: SBa, SBb, and SBc; are the intermediate between spiral and elliptical, they are labeled S0 or SB0, it consists of a very bright bulge with no surrounded arms and that what makes it difference from an E0 galaxy [6].  Irregular galaxies: Im, and Ibm, some galaxies don't have neither a spiral structure nor a nuclear bulge, they appear as a random collection of stars with no obvious order.…”

“…Principle Component Analysis was used to reduce the dimensionality of the data, and to extract relevant information in the images. [6], compared and evaluated the behavior of 10 artificial neural networks based classifiers based on selected sets of galaxy features derived from image analysis and principle component analysis features (PCA). The support vector machine (SVM) based classifier provides the best results about 99.5%.…”