The extraction algorithm for LAMOST two-dimensional spectrum image
Abstract
Spectra extraction is an important procedure in reducing raw data of fiber-fed spectrographs. The flux of fiber on each wavelength will be obtained by extracting spectra from a two-dimensional image of fiber-fed spectrographs. Spectra extraction affects the later spectrum data processing procedures directly, and its accuracy guarantees the scientific value of spectra. The extraction algorithm is described for extracting one-dimensional object spectra from a two-dimensional spectrum image of LAMOST in this paper. The principle of spectra extraction based on the weighted least square method and polynomial fitting is presented in detail as well as fiber spatial profile. The raw data of fiber-fed spectrographs consist of a collection of spectra distributed along a certain axis of a two-dimensional frame. For each wavelength, spectrum expands along the cross-dispersion axis and covers some width, which is called spatial profile. The profile is considered Gassian function in the paper. According to Guassian spatial profile character of the fiber and experiments, the scope of sampling point is well selected in this paper. The good scope of sampling point solves the problem that the fibers' flux extracted is negative. The least square method is often used for scientific computing. To fit Guassian spatial profile of fiber the weighted least square method is used for spectra extraction of LAMOST, and then the flux of fiber on each wavelength is obtained further. The existence of noise is usually a problem in data processing. In reducing raw data of fiber-fed spectrographs, strong noise will distort fiber spatial profile seriously. It will lead the bad extraction results from the spectrum image and can't meet the scientific demand. In frequency domain, the spectra energy lies in the low frequency, but the noise energy lies in the high frequency on the contrary. So the improved extraction algorithm based on spectrum analysis in frequency domain is proposed against the affection of strong noise in this paper. In the first step, it filters the sharp noise out by Fast Fourier Transformer and low-pass filter, and gets a more accurate spatial profile, which has less affection of noise. Then the spectra are extracted with existing algorithm in the later step. The cut-off frequency in improved spectrum extraction is a fatal parameter, and the iteration algorithm for selecting cut-off frequency is described in detail. Finally the improvement is tested with the simulated data provided by LAMOST 2d-pipeline system, the results demonstrate the improved spectrum extraction can remove sharp noise effectively and does not twist the original spatial profile. More accurate extraction results are also achieved. Furthermore, the improvement in extreme cases of 5.6 magnitudes difference between neighborhood fibers is tested, and the results also demonstrate the feasibility and effectiveness of the proposed method.
- Publication:
-
Acta Astronomica Sinica
- Pub Date:
- July 2008
- Bibcode:
- 2008AcASn..49..327C
- Keywords:
-
- Telescopes: Techniques: Spectroscopic