1 Introduction

The analysis of the distribution of spectral types in stellar systems is a powerful diagnostic for the estimation of their composition, age and evolutionary stage. The observing material mainly used for this task are prism spectral plates taken with Schmidt-class telescopes. Such a plate generally contains thousands of spectra, and there are prism-plate libraries and digitized data bases in several astronomical centers that can be exploited for this analysis. To deal with the pool of spectral data and to fully exploit them, highly automated image analysis tools need to be developed.

Extracting the physical quantities from the digitized spectral plates involves three main stages: detection of the spectra, extraction of their images, and classification of the spectra. The purpose of this paper is to present a new, fully automated method for the detection of spectra. Previous works for automated classification of stellar spectra (von Hippel et al. 1994), for quasar spectral analysis (Hewett et al. 1995), for galaxy classification (Lahav et al. 1996) or for galaxy redshift measurements (Tucholke & Schuecker 1992; Schuecker 1996) applied spectra detection using coordinates of the corresponding stars either determined for the purpose with the Automated Plate Measuring (APM) system or taken from catalogues (e.g. GSC). Our method is based on processing only the digitized prism plates without the need of the corresponding direct plates, necessary for the above-mentioned methods.

For this study, high-quality film copies of IIIa-J plates taken with the 1.2 m UK Schmidt Telescope in Australia have been used. The spectral plates are with dispersion of 830 Å/mm at H$_{\gamma}$ and spectral range from 3400 to 5000 Å. The magnitude limit is about 16.5 in V. The photographic material has been digitized at the Trieste Observatory by means of a PDS 1010A microdensitometer and at the Royal Observatory of Edinburgh using the Super-COSMOS measuring machine. The spectra detection method has been developed under MIDAS (Midas 96NOV 1996) as a part of the objective prism plate image processing context OBJPR (Pasian et al. 1997).

The new algorithm, based on signal processing methods, is presented in Sect. 2, and its implementation is described in Sect. 3. Results from the tests of the new method are given in Sect. 4.