天体光谱的自动处理算法研究

发布时间：2018-02-11 02:50

本文关键词： 超新星候选光谱分类局部均值质量估计　出处：《辽宁科技大学》2015年硕士论文　论文类型：学位论文

【摘要】：郭守敬望远镜,(简称LAMOST)是国家“九、五”重大科学工程项目之一,是一架横卧于南北方向的新类型的大视场兼备大口径的中星仪式反射施密特望远镜。LAMOST在每个观测夜能够获得多达几万条光谱数据,是目前世界上光谱获取率最高的望远镜。面对LAMOST所产生的海量光谱数据,一些传统的人工处理光谱的方法已不再使用,迫切需要研究和寻找一些高效率的光谱自动处理方法。本文研究的工作主要包括以下三部分:(1)超新星候选范围的快速约减。本文提出了一种新的超新星候选范围约减算法(SKLOF),根据超新星本身具有稀有的特性,首先利用数据剪枝思想剪枝掉大部分肯定不是超新星候选的光谱数据,然后利用PCA算法对光谱数据进行分析并构造超新星特征空间,利用改进后的LOF算法对光谱数据进行离群分析。(2)天体光谱的分类。本文利用SDSS-DR8的光谱数据,将基于局部均值的K近质心近邻点(Local Mean-based K-Nearest Centroid Neighbor,LMKNCN)算法应用到恒星、星系和类星体的分类中,并且取得了较好地效果。由于星系和类星体都属于河外天体,红移比较大,二者的光谱分类在光谱自动处理中具有一定的代表性,因此本文同时也研究了基于局部均值的K近邻算法在星系和类星体光谱的分类中的应用,并且取得了较高的光谱分类正确率。(3)恒星光谱大气物理参数测量。恒星大气参数即表面有效温度(Teff)、表面重力加速度(log g)以及金属丰度([Fe/H])对科学家研究宇宙的演化等重大科学问题具有重要的意义。因此本文研究了基于质量估计的恒星大气物理参数自动测量方法,该方法首先是建立质量分布,将原始光谱数据经过质量估计算法映射到质量空间,然后在质量空间利用SVR方法对恒星的大气物理参数进行预测且取得了不错的效果。
[Abstract]:Guo Shoujing Telescope (LAMOST) is one of the major scientific engineering projects in the Ninth Five-Year Plan. It is a new type of large field of view lying in the north-south direction, with a large aperture, medium Star ritual reflection Schmidt Telescope. LAMOST can obtain up to tens of thousands of spectral data per observation night. Is the world's highest spectral acquisition rate of telescopes. In the face of the massive spectral data generated by LAMOST, some traditional manual spectral processing methods are no longer used. There is an urgent need to study and find some efficient automatic spectral processing methods. The work of this paper mainly includes the following three parts: 1) the rapid reduction of the candidate range of supernova. In this paper, a new candidate range of supernova is proposed. According to the rare characteristics of supernova itself, Firstly, the data pruning idea is used to prune most of the spectral data which is definitely not a supernova candidate. Then the spectral data are analyzed by using PCA algorithm and the supernova feature space is constructed. The improved LOF algorithm is used to classify the spectral data of outliers. In this paper, based on the spectral data of SDSS-DR8, the local Mean-based K-nearest Centroid neighbor LMKNCNC algorithm based on local mean is applied to stars. In the classification of galaxies and quasars, good results have been obtained. Since galaxies and quasars belong to extra-river bodies and have a relatively large redshift, their spectral classification is representative in the automatic processing of spectra. Therefore, this paper also studies the application of K-nearest neighbor algorithm based on local mean in the spectral classification of galaxies and quasars. A high spectral classification accuracy rate of spectral classification has been obtained. The measurements of the spectral atmospheric physical parameters of stars are obtained. The stellar atmospheric parameters, I. e., surface effective temperature, surface gravity acceleration, log g) and metal abundance ([Fe/H]) are used to study the evolution of the universe. Therefore, an automatic measurement method for atmospheric physical parameters of stars based on mass estimation is studied in this paper. The first step of this method is to establish the mass distribution and map the original spectral data to the mass space through the quality estimation algorithm. Then the SVR method is used to predict the atmospheric physical parameters of stars in the mass space and good results are obtained.
【学位授予单位】：辽宁科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P111.23

【相似文献】