Share Email Print

Proceedings Paper

New automated classification technique of galaxy spectra with Z<1.2 based on PCA-ODP
Author(s): Dongmei Qin; Zhanyi Hu; Yongheng Zhao
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

In this paper, we investigate the Principal Component Analysis-Optimal Discrimination Plane (PCA-ODP) approach on a data set of galaxy spectra including eleven standard subtypes with the redshift value ranging from 0 to 1.2 and a span of 0.001. These eleven subtypes are E, S0, Sa, Sb, Sc, SB1, SB2, SB3, SB4, SB5, SB6, respectively, according to the Hubble sequence. Among them, the first four subtypes belong to the class of normal galaxies (NGs); the remaining seven belong to active galaxies (AGs). We apply the PCA approach to extract the features of galaxy spectra, project the samples onto the PCs, and investigate the ODP method on the data of feature space to find the optimal discrimination plane of the two main classes. ODP approach was developed from Fisher's linear discriminant method. The difference between them is that Fisher's method uses only one Fisher's vector and ODP uses two orthogonal vectors including Fisher's vector and another. Besides the data set above, we also use the Sloan Digital Sky Survey (SDSS) galaxy spectra and Kennicutt (1992) galaxy data to test the ODP classifier. The experiment results show that our proposed technique is both robust and efficient. The correct rate can reach as high as 99.95% for the first group data, 96% for SDSS data and 98% for Kennicutt data.

Paper Details

Date Published: 19 December 2002
PDF: 9 pages
Proc. SPIE 4847, Astronomical Data Analysis II, (19 December 2002); doi: 10.1117/12.460379
Show Author Affiliations
Dongmei Qin, National Lab. of Pattern Recognition (United States)
Zhanyi Hu, National Lab. of Pattern Recognition (United States)
Yongheng Zhao, National Astronomical Observatories (United States)

Published in SPIE Proceedings Vol. 4847:
Astronomical Data Analysis II
Jean-Luc Starck; Fionn D. Murtagh, Editor(s)

© SPIE. Terms of Use
Back to Top