We analyze deep HST/WFPC2 images in U,B,V,I using artificial neural network (ANN) classifiers which use as input i) global photometric parameters, and ii) Fourier coefficients used to construct two-dimensional galaxy image models. These ANN classifiers distinguish quite well between E/S0's, Sabc's, and Sd/Irr+M (M for merging systems) for B<=27 mag. We present classifications in UBVI from (a) independent human classifiers; (b) from ANN's trained on V and I images; and (c) from an ANN's trained on images in the rest-frame UBV according to the expected redshift distribution as a function of B magnitude. For each of the three methods, we find that the fraction of galaxy types does not significantly depend on wavelength, and they produce consistent counts as a function of type. We compare the morphological properties of our high redshift HST-observed galaxies to those for local samples and present the B-band galaxy counts derived from WFPC2 fields as a function of morphological type. E/S0's are only marginally above the no-evolution predictions, and Sabc's are at most 0.5 dex above the non-evolving models for B<=24 mag. The faint blue galaxy counts in the B-band are dominated by Sd/Irr+M's, and can be explained by a moderately steep local LF undergoing strong luminosity evolution.
American Astronomical Society Meeting Abstracts
- Pub Date:
- December 1996