Automated Recognition of Galaxy Morphology Using Neural Networks

Previous abstract Next abstract

Session 107 -- Galaxies: Morphology and Interactions
Display presentation, Thursday, 12, 1995, 9:20am - 6:30pm

[107.09] Automated Recognition of Galaxy Morphology Using Neural Networks

Nielsen, M.L. and Odewahn, S.C. (University of Minnesota)

The feasibility of using neural networks to recognize morphological features of galaxies is examined. Neural networks are an artificial intelligence technique which simulate groups of biological neurons and their interconnections to utilize their ability to learn and generalize. The networks are trained with a set of galaxies of known morphology imaged in the O and E bandpasses from the region of the North Galactic Pole. The galaxy images were digitized by the Minnesota Automated Plate Scanner from nine fields of the first epoch Palomar Observatory Sky Survey. Once trained, the networks are tested with an independent set of galaxies from the same field to assess their ability to recognize features such as spiral structure and bars. The use of image data as direct inputs to the neural networks is compared to that of using photometric properties derived from the images.

Thursday program listing