[Previous] | [Session 20] | [Next]
D.D. Proctor (LLNL)
It has been shown that bent-double radio sources can act as tracers for clusters of galaxies (Blanton, et al., 2000, ApJ 531, 118). Results of work to automate selection of bent-double sources from the FIRST (Faint Images of the Radio Sky at Twenty Centimeters) Survey database using pattern recognition techniques are presented. The task is complicated by the low resolution (in a pattern recognition sense) of the images, chance superposition of sources, the ambiguous nature of visual classifications and considerable variation in bent morphology. Decision tree and artificial neural network methods are compared and show substantially equivalent results. Resulting classifiers provide well-defined samples for future studies.
We acknowledge computing and office facilities provided by IGPP/LLNL and computer resources provided by the FIRST project.