[Previous] | [Session 77] | [Next]
S. Odewahn, S. Cohen, R. Windhorst (Arizona State University)
We present a morphological approach to automated galaxy classification based on the Fourier reconstruction of galaxy images and subsequent pattern analysis based on the amplitude and phase angle of the Fourier components used. This method quantitatively detects the presence of and characterizes the properties of spiral arms, bars and rings. In addition, it provides a systematic means of describing the degree of large-scale global asymmetry in a galaxy, a property known to correlate loosely with Hubble stage, but one which is also strongly linked to important physical events such as merging and tidal interaction. We describe the method and discuss the role of systematic and accidental errors in its use caused by varying image resolution and signal-to-noise. We compare independent sets of visual estimates of morphological types in the revised Hubble system to establish a well understood set of training/testing cases. This is comprised of a local sample from ground-based imaging and a distant sample collected from the HST archive. These data are used to train artificial neural network classifiers that use Fourier amplitude and phase information as the input pattern information. This method will be used to study the morphological evolution of galaxies in the high-redshift regime. This work was funded under NASA grant AR.7534.01.96A.