The Research On Several Issues Of The Two Dimensional Fiber Spectrum Data Processing

As the development of the astronomical observation and fiber technics, kinds of fiber spectral telescope are enjoyed by astronomers from different countries because of their large field, good observation depth and high efficiency. Several fiber spectral telescopes are founded in the world recently, such as the SDSS of America, the AAT of Australia, and the LAMOST of China. These telescopes can receive and imaging multiple target spectra in the same time on the CCD through exposure, because they have multiple fibers.The processing of 2-D fiber spectral data is one of the key ways to obtain the information of target stars by astronomers through the observed data of fiber telescopes. The precision of data processing will affect the accuracy and integrality of observed spectral information. There are many effective methods of data processing in information domain. If these methods can be applied to astronomical data processing by combining the characteristics of spectral data, good results will be achieved, and it has important significance of research and application value.First, the mechanism of 2-D fiber spectral images is introduced in brief, then the flow path of data processing is analyzed based on the 2-D fiber spectral model and characteristics of LAMOST and SDSS. On this basis, the detection and rejection of cosmic rays on the single CCD image, the information recovery of fiber profiles in the low SNR condition, the profile fitting of non-Gaussian distribution, and the spectrum extraction based on the 2-D Gaussian model are the focal points of this paper. The main work and innovations are listed as follows:The detection and rejection of cosmic rays on the single CCD image is researched. First the different formation mechanisms between the cosmic rays and the spectral profiles are introduced. Based on the characteristics and differences, the double threshold algorithm is employed to determine the suspected cosmic ray point set; then the Bessel curve fitting algorithm of the graphics is used to confirm the location of cosmic rays, the neighborhood average value method is applied to recover the original value which is polluted by cosmic ray. Experimental results show that the proposed method, detection and rejection of cosmic rays on single CCD image can detect the location of cosmic rays well, and have high detection probability and meet the requirements of the follow-up treatment. For the different images with the same full width at half maximum the threshold of this method is not needed to be changed, and the algorithm has good stability.The information extraction of fiber profiles under the conditions of low SNR is researched. If the observation targets have high magnitude (ie, the target stars are faint ones), the observed 2-D fiber spectral data have so low SNR that the fiber profiles are buried by noise. Profile fitting method can not be directly used to extract the spectra of these data. Then empirical mode decomposition and stationary wavelet decomposition are employed to analyse the observational data and the corresponding flat-field image is used as a reference image. Finally, the data are reconstructed after filtering to recover the fiber profiles polluted by noise. The experimental results show that the proposed methods can well recover the shapes and characteristics of the fiber profiles, and the recoved profiles can meet the fiber profile requirement of the follow-up high-precision spectrum extraction algorithm.The profile fitting for the non-Gaussian distributional fibers is researched. Because the effects of optical grating and CCD camera performances, the fiber profiles by exposure are not the normative Gaussian distribution but with asymmetric or local flattened topfeatures. The full width at half maximum (FWHM) of point spread function in LAMOST is relatively wider than other telescopes, so this effect is more pronounced. First, the differences between the observed fiber profiles and the theoretical Gaussian model profiles are introduced based on the observed data of LAMOST. Then, double exponential model is taken to fit the observed fiber profiles. Experimental results show that for the data which satisfy the requirement of the signal to noise ratio the proposed method can achieve good fitting results. Especially for flattened top and asymmetrical profiles, the approach can reach good results which are very close to the real ones.The extraction problem based on the two 2-D Gaussian model is researched. The traditional extraction spectrum methods only use the 1-D model to extract the fiber spectra based on the 1-D data in the spatial direction of the 2-D image. In fact, the target signal is projected on the CCD through the grating and other equipments as the form of 2-D point spread function of the pixel to be diffused to its neighborhood area for each fiber center, and 1-D extraction spectrum methods ignore this feature. Proposed 2-D spectrum extraction method makes use of the 2-D characteristics of the adjacent area data, the extracted spectra have more integrated information and more resistance to noise. The characteristics and effectiveness are demonstrated by experiments.In summary, the paper researches several issues of the 2-D fiber spectral data processing, and the corresponding algorithms are proposed. These contribute to the accuracy improving of the 2-D fiber spectra data processing, and have significances for the developments of the 2-D fiber spectra data processing. For the data processing of the LAMOST and other fiber spectral telescopes, the researches in this paper also have certain application value and the use for reference.