Abstract：Quasar Objects (QSOs) are detectable at very large distance,with broad,red-shifted emission lines,strong ultraviolet and strong time variability of the optical light.QSOs play an important role in the research of the universe.The main purposes of quasar recog-nition are to identify the emission peaks in an observable quasar spectrum and to determine the observable quasar’s redshift value.Due to the inherent extremely noisy characteristics of quasar spectrums and the limitation of observable conditions,automatic quasar recognition is a hard problem to tackle,and the commonly used direct matching approaches based on rules are ineffective.This paper introduces a stratified approach based on Hough transform and neural network which is shown to be simple,efficient,robust and easy to generalize.
2000, Vol. 22 
