The Research On Some Key Technologies Of The Two Dimensional Fiber Spectrum Data Processing

Astronomical observation techniques have been very different from the past years. Especially, the multi-object fiber spectroscopic technique becomes more and more mature, and it has been widely adopted in astronomical research because of its wide field of view, large observation depth and high efficiency. The Guoshoujing Telescope, also called the Large Sky Area Multi-Object Fiber Spectroscopic Telescope (LAMOST), is a newly constructed astronomical system in China, which is designed based on the multi-object fiber spectroscopic technique. LAMOST can observe almost4000celestial objects at one time and it has a very high spectrum acquiring rate.While the multi-object fiber spectroscopic technique improves the efficiency of astronomical observation, it also increases the complexity of the telescope system integration. Massive growth of astronomical observation data make that it needs more accurate astronomical data processing algorithm. The two-dimensional fiber spectrum data processing is an important part of the astronomical data processing. It directly affects the follow-up spectral analysis and celestial target recognition, and it is an important way to the astronomical community to obtain various kinds of astrophysical information. Therefore, considering the features of the two-dimensional fiber spectrum data processing, combined with a variety of innovative technology in the field of information processing, researching how to improve the accuracy and reliability of astronomical data processing has important research significance and application value.Based on the two-dimensional fiber spectrum data model of the GuoShouJing Telescope system, this thesis first gave a brief description of the operating principle of telescope system and the generation procedure of spectrum data. Then a simple analysis of the conventional process flow based on the data generated by the LAMOST system was given. Based on the analysis of the spectrum process flow, several key technologies was selected as the focal points of this thesis, including the detection and elimination of cosmic-rays on the spectral CCD image, high-precision sky subtraction, and the detection of fiber profiles in low signal-to-noise ratio (SNR) condition.The main work and innovations are listed as follows: The detection and elimination of cosmic-rays in the single exposure spectral CCD images was studied. Based on the research and analysis of the different imaging mechanism on the CCD between cosmic-ray hits and spectra, we took full advantage of their differences in energy value and the shape of profile. The algorithm based on gray-scale mathematical morphology was applied for the detection and elimination of cosmic-rays. For the two-dimensional fiber spectrum CCD image observed in single exposure, the operations of erosion and dilation were used to deal with images successively, with appropriate structuring elements. Points of cosmic-ray hits can be detected by analyzing the variation between data pre-and post-operation. Selecting an appropriate morphological structuring element for processing can recover the value of the pixels polluted by cosmic-ray hits. Experiments were based on both simulated LAMOST images and observed SDSS images. The results showed that the algorithm can detect the location of cosmic-ray hits in the image accurately. Compared to the traditional algorithm, the detection efficiency in our algorithm is raised and the probability of false alarm decline.The high-precision sky subtraction algorithm was researched. Through research we found that the sky subtraction is performed after the spectrum extraction in the traditional two-dimensional fiber spectra data processing flow. The two-dimensional spectrum becomes a one-dimensional data after the spectrum extraction processing. Superposition of noise and accuracy of the spectrum extraction processing have a direct impact on the treatment effect of sky subtraction. To take full advantage of the information of the two-dimensional sky spectral data before the spectrum extraction processing, a novel sky subtraction algorithm based on two-dimensional sky-background modeling was proposed. Before the spectrum extraction processing, the two-dimensional sky-background model was synthesized by using B-spline curve fitting method and two-dimensional sky sampling data. Then the two-dimensional sky-background can be subtracted from the object spectrum according to the wavelength position in order to achieve the sky subtraction processing. The experimental results show that, compared to the traditional approach, the sky component can be removed more thoroughly in this algorithm and the residual spikes in the object spectrum after sky-subtraction processing are smaller and smoother. This algorithm can restore the original object spectrum information better and it meets the requirements of the spectrum quality for the follow-up spectral analysis.The profile detection method for astronomical spectrum data with low SNR was researched. When the observed celestial objects have high magnitude, ie the observed targets are dark celestial bodies, the obtained two-dimensional fiber spectrum data have very low SNR. For this kind of spectrum data, the spectrum profiles are polluted so heavily that the centers of profiles deviate from the ideal locations, and the shapes of contours are not smooth and regular anymore, which makes that the profile fitting method can not be performed directly for the spectrum extraction processing. Therefore, it is necessary to perform this pretreatment for two-dimensional astronomical spectrum data before the extraction of spectra. The research is based on the time-frequency analysis method. The different features between spectrum signal and noise in different time-frequency transforms were analyzed according to their distribution characteristics in the spatial direction. Two time-frequency analysis methods, included the Wigner Transform and the Wigner Bispectrum, were adopted for the analysis of observed spectrum data respectively. The corresponding flat-field signal was employed as a reference signal to obtain the cut-off frequencies of filters. Then filtering was performed for the transformed signal. The detected profile signal can be obtained by using the reconstruction method. Experiments with both simulated and observed data based on the LAMOST project were presented to demonstrate the effectiveness of the proposed method. It indicates that the proposed algorithms can well extract the original spectrum profile information which is destroyed by noise, and the detected profiles can meet the requirements of the quality of fiber contour for the subsequent high-precision spectrum extraction processing.