METHODS FOR UNSUPERVISED ARC-LINE IDENTIFICATION

Four different methods for automated identification of emission lines in calibration (are) spectra are discussed. These methods can be applied in batch mode, because they do not require interactively entered information, and may find application in pipe-line processing of spectroscopic observations. Input data for the algorithms consists of a table of (pixel) positions of detected are-lines, and a reference list of wavelengths of standard spectral lines. A pattern-recognition based method requires accurate spectral dispersion information in addition. Other methods, however, can be used in a search mode, in order to determine the dispersion from the data themselves. These methods exhibit a varying degree of robustness in cases of unfavourably conditioned input information, which makes some methods more suitable for a particular situation than the others.