The Profile Recovery And Fitting Method Of The2-D Fiber Spectral Data With Low SNR

The LAMOST (Full name is Large Sky Area Multi-Object Fiber Spectroscopy Telescope), also known as Guoshoujing Telescope, is a self-developed reflecting Schmidt telescope of China which has both large aperture and large field of view. Its establishment makes China march toward a new stage in the technical field of optical instrument manufacturing and mass information processing. LAMOST also provides an excellent observation tool and research platform for the observation and study of astronomy in the world.Based on LAMOST telescope data processing system, this thesis describes the type of the observed data and the process flow of the subsystem. Then the work focus on the preprocessing before spectrum extraction. As the two dimensional observed spectral data have low signal to noise ratio (SNR), we need to recover the spectral profile to improve the quality of spectral image, including the detection and removal of cosmic rays and noise reduction of the spectral signal. As the spectral profiles have asymmetric or top-flattened feature, a new profile fitting model and iterative solution is proposed to improve the fitting precision preparing for the high accuracy spectrum extraction. The main contributions in the thesis are described as follows:1. The method of cosmic-ray detection and removal is studied. According to the difference of cosmic rays and the spectral outlines in shape and gray values in the image, a cosmic-ray detection approach based on the local information active contour model is proposed. The global optimal solution is obtained quickly using a dual form of the TV-norm. Then image restoration model is used to remove ray points. Experimental results exhibit that our method is right, this step makes preparations for the denoising of low SNR spectral signal.2. The denoising method for the one-dimensional spectral signal with low SNR is researched. As the traditional low-pass filter can not filter out aliasing noise, a new weighting filter based on priori SNR in Wigner transform domain is proposed, according to the distribution difference of spectral signal and noise in the time-frequency domain. An appropriate weighting function is designed which can suppress aliasing noise while maintaining effective signal. The results of simulated and observed data indicate that the denoising results can meet the need of subsequent spectrum extraction. 3. The spectral profile fitting of non-Gaussian model is studied. As poor adaptability of existing models and limitation of general solution methods, a multi-Gauss-Lorenz curve superimposed model is proposed to approximate spectral profiles, then Nelder-Mead simplex iterative method is used to optimize the objective function. It can flexibly adapt to the changes in the profile shape, and has high fitting precision and a certain anti-jamming ability. Experiments on simulated and observed data indicate that our method is valid.