Automated analysis of objective-prism spectra – I. Quasar detection

A fully automated system for the location, measurement and analysis of large numbers of low-resolution objective-prism spectra is described. The system is based on the APM facility at the University of Cambridge, and allows processing of objective-prism, grens or grism data. Particular emphasis is placed on techniques to obtain the maximum signal-to-noise ratio from the data, both in the initial spectral estimation procedure and for subsequent feature identification. Precise wavelength scales for each spectrum and the definition of magnitude-limited samples simplify subsequent analysis considerably. Comparison of a high-quality visual catalogue of faint quasar candidates with an equivalent automated sample demonstrates the ability of the APM system to identify all the visually selected quasar candidates. In addition, a large population of new, faint ($m_J \sim 20$) candidates is identified. Subsequent slit spectroscopy confirms the high quasar detection success rates of both the visual and automated samples. The APM system is the first to select large homogeneous quasar samples, with surface densities of ∼20 per square degree to $m_J \simeq 20$ in a fully automated fashion, with practically no visual inspection of candidates necessary. No calibration information whatsoever is required, although such information may be readily incorporated if available. The advantages of an object selection philosophy which involves searching for all unusual objects, rather than objects of a specified type are stressed.