Object detection is a common problem in image processing and particularly in astronomy and astrophysics, when a huge number of objects are to be processed. Such is the case of objective prism plates, where thousands of spectra are recorded. We present a new method for automatic detection of stellar spectra on digitally-scanned objective prism plates. Our method is based on signal processing techniques with emphasis on obtaining the maximum signal-to-noise ratio from the data, as well as on

more »

... ing a number of morphological criteria of one-dimensional stellar spectra images. It uses only the prism plate image without the need of the corresponding direct plate. This fully automated method gives inpressive results in the detection of stellar spectra.