LAMOST恒星大氣參數(shù)提取系統(tǒng)
發(fā)布時(shí)間：2018-10-16 15:48
【摘要】： 隨著LAMOST銀河系巡天計(jì)劃的開展,每個(gè)觀測(cè)夜將獲得上萬(wàn)條恒星光譜。光譜蘊(yùn)含著天體的重要信息,通過(guò)恒星光譜來(lái)得到恒星的大氣物理參數(shù)是天文學(xué)中的一個(gè)基礎(chǔ)工作,因此恒星光譜分析在天體研究中占有重要地位。通過(guò)恒星光譜快速、準(zhǔn)確、自動(dòng)的提取恒星大氣物理參數(shù)是非常值得研究和探索的。本研究針對(duì)LAMOST的需求,設(shè)計(jì)、實(shí)現(xiàn)了一套恒星大氣參數(shù)提取系統(tǒng)。主要研究工作如下： 1、針對(duì)LAMOST的觀測(cè)光譜進(jìn)行預(yù)處理,利用11條強(qiáng)吸收線的觀測(cè)波長(zhǎng)和實(shí)驗(yàn)室波長(zhǎng)的對(duì)比,計(jì)算得到視向速度,對(duì)光譜進(jìn)行視向速度校正；然后對(duì)光譜的藍(lán)段(3850-6000 A)和紅端(6000-9000 A)分別進(jìn)行多項(xiàng)式擬合,然后再綜合進(jìn)行多項(xiàng)式擬合,提取全局連續(xù)譜；針對(duì)83條原子線和分子線進(jìn)行譜線特征提取等。 2、利用網(wǎng)格模板匹配提取恒星大氣參數(shù)。使用Kurucz模型生成覆蓋網(wǎng)格節(jié)點(diǎn)的兩套理論光譜模板,一套為包含g-r色指數(shù)和4400-5500 A的標(biāo)準(zhǔn)化光譜,一套只包含4400-5500 A的標(biāo)準(zhǔn)化光譜。定義觀測(cè)光譜和理論模板光譜之間的距離,利用Nelder-Mead算法快速搜索極小值,利用最接近的理論光譜的參數(shù)作為觀測(cè)光譜的恒星大氣參數(shù),最后利用蒙特卡洛模擬噪聲的分布,得到恒星大氣參數(shù)的誤差。 3、使用PCA降維的恒星光譜數(shù)據(jù)作為輸入,利用神經(jīng)網(wǎng)絡(luò)提取恒星大氣參數(shù)。將光譜的紅藍(lán)端分別降到二十五維,作為神經(jīng)網(wǎng)絡(luò)的輸入,三個(gè)恒星大氣參數(shù)作為輸出,中間隱藏節(jié)點(diǎn)為十個(gè),構(gòu)建三層神經(jīng)網(wǎng)絡(luò)。使用理論光譜和SLOAN光譜(使用SSPP測(cè)量參數(shù))作為訓(xùn)練數(shù)據(jù)及測(cè)試數(shù)據(jù),訓(xùn)練得到兩套神經(jīng)網(wǎng)絡(luò)系統(tǒng)。 4、使用卡方最小化技術(shù)提取恒星大氣參數(shù)。首先生成兩套不同的理論光譜模板,定義觀測(cè)光譜和理論光譜之間的卡方距離,為了減少計(jì)算量,利用半流量點(diǎn)技術(shù)來(lái)進(jìn)行初始的溫度估計(jì),然后使用剪枝的多項(xiàng)式擬合技術(shù)得到最小值,求得有效溫度,使用同樣的步驟依次求得表面重力值和金屬豐度值。第二套模板中使用第一套模板求得得有效溫度,不過(guò)第二套模板將在以后的應(yīng)用中計(jì)算alpha元素豐度。在本系統(tǒng)中,我們還實(shí)現(xiàn)了通過(guò)觀測(cè)的g-r色指數(shù)和通過(guò)巴爾默(Blamer)線系的強(qiáng)度預(yù)測(cè)得到有效溫度,最終使用了兩個(gè)理論的有效溫度估計(jì)和三個(gè)經(jīng)驗(yàn)有效溫度估計(jì)。 5、利用銀河系中的球狀星系團(tuán)和疏散星團(tuán)的金屬豐度值對(duì)本系統(tǒng)的參數(shù)值準(zhǔn)確性進(jìn)行了評(píng)估,并使用其他望遠(yuǎn)鏡觀測(cè)的高分辨率光譜提取的參數(shù)作為真實(shí)值,對(duì)本系統(tǒng)中的金屬豐度參數(shù)進(jìn)行了校正,得到每個(gè)算法在不同區(qū)間的誤差和彌散度,對(duì)結(jié)果進(jìn)行了重新加權(quán),獲得了較好的準(zhǔn)確性。
[Abstract]:With the launch of the LAMOST Galactic Sky Survey program, tens of thousands of star spectra will be obtained per observation night. Spectra contain important information of celestial bodies. It is a basic work in astronomy to obtain the atmospheric physical parameters of stars by stellar spectra, so star spectral analysis plays an important role in the study of celestial bodies. It is worth studying and exploring to extract the atmospheric physical parameters of stars quickly, accurately and automatically through star spectrum. In order to meet the requirements of LAMOST, a stellar atmospheric parameter extraction system is designed and implemented in this paper. The main research work is as follows: 1. The observation spectrum of LAMOST is pretreated, and the apparent velocity is calculated by comparing the observation wavelength of 11 strong absorption lines with the wavelength of laboratory, and the spectrum is corrected by apparent velocity. Then the blue region (3850-6000A) and the red end (6000-9000 A) of the spectrum are fitted by polynomial respectively, and then the global continuous spectrum is extracted by comprehensive polynomial fitting. For 83 atomic and molecular lines, the spectral line features are extracted, etc. 2. The atmospheric parameters of stars are extracted by mesh template matching. The Kurucz model is used to generate two sets of theoretical spectral templates covering the grid nodes, one set of standardized spectra containing g-r color index and 4400-5500A, and one set containing only 4400-5500A standardized spectra. The distance between the observed spectrum and the theoretical template spectrum is defined. The Nelder-Mead algorithm is used to quickly search the minimum, and the nearest theoretical spectrum parameters are used as the atmospheric parameters of the observed spectrum. Finally, Monte Carlo is used to simulate the distribution of noise. The errors of stellar atmospheric parameters are obtained. 3. The star atmospheric parameters are extracted by neural network using PCA reduced dimension star spectral data as input. The red and blue side of the spectrum is reduced to 25 dimension, which is used as the input of the neural network, three star atmospheric parameters as the output, and ten hidden nodes in the middle. The three-layer neural network is constructed. Using theoretical spectrum and SLOAN spectrum (using SSPP measurement parameters) as training data and test data, two sets of neural network systems are trained. 4. Chi-square minimization technique is used to extract stellar atmospheric parameters. First, two sets of different theoretical spectral templates are generated to define the chi-square distance between the observed spectrum and the theoretical spectrum. In order to reduce the computational complexity, the half-flow point technique is used to estimate the initial temperature. Then the pruning polynomial fitting technique is used to obtain the minimum value and the effective temperature, and the surface gravity value and the metal abundance value are obtained in turn by the same steps. The first set of templates is used to obtain the effective temperature in the second set of templates, but the second set of templates will be used to calculate the abundance of alpha elements in future applications. In this system, we also achieve the effective temperature obtained by the observed g-r color index and the strength prediction of the Balmer (Blamer) line system. Finally, two theories of effective temperature estimation and three empirical effective temperature estimates are used. 5. The accuracy of the system parameters is evaluated by using the metallic abundance values of the globular clusters and the open clusters in the Milky way. Using the parameters extracted from high-resolution spectra observed by other telescopes as real values, the parameters of metal abundance in this system are corrected, the errors and dispersion of each algorithm in different regions are obtained, and the results are reweighted. Good accuracy has been obtained.
【學(xué)位授予單位】：山東大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2010
【分類號(hào)】：P144
本文編號(hào)：2274839
[Abstract]:With the launch of the LAMOST Galactic Sky Survey program, tens of thousands of star spectra will be obtained per observation night. Spectra contain important information of celestial bodies. It is a basic work in astronomy to obtain the atmospheric physical parameters of stars by stellar spectra, so star spectral analysis plays an important role in the study of celestial bodies. It is worth studying and exploring to extract the atmospheric physical parameters of stars quickly, accurately and automatically through star spectrum. In order to meet the requirements of LAMOST, a stellar atmospheric parameter extraction system is designed and implemented in this paper. The main research work is as follows: 1. The observation spectrum of LAMOST is pretreated, and the apparent velocity is calculated by comparing the observation wavelength of 11 strong absorption lines with the wavelength of laboratory, and the spectrum is corrected by apparent velocity. Then the blue region (3850-6000A) and the red end (6000-9000 A) of the spectrum are fitted by polynomial respectively, and then the global continuous spectrum is extracted by comprehensive polynomial fitting. For 83 atomic and molecular lines, the spectral line features are extracted, etc. 2. The atmospheric parameters of stars are extracted by mesh template matching. The Kurucz model is used to generate two sets of theoretical spectral templates covering the grid nodes, one set of standardized spectra containing g-r color index and 4400-5500A, and one set containing only 4400-5500A standardized spectra. The distance between the observed spectrum and the theoretical template spectrum is defined. The Nelder-Mead algorithm is used to quickly search the minimum, and the nearest theoretical spectrum parameters are used as the atmospheric parameters of the observed spectrum. Finally, Monte Carlo is used to simulate the distribution of noise. The errors of stellar atmospheric parameters are obtained. 3. The star atmospheric parameters are extracted by neural network using PCA reduced dimension star spectral data as input. The red and blue side of the spectrum is reduced to 25 dimension, which is used as the input of the neural network, three star atmospheric parameters as the output, and ten hidden nodes in the middle. The three-layer neural network is constructed. Using theoretical spectrum and SLOAN spectrum (using SSPP measurement parameters) as training data and test data, two sets of neural network systems are trained. 4. Chi-square minimization technique is used to extract stellar atmospheric parameters. First, two sets of different theoretical spectral templates are generated to define the chi-square distance between the observed spectrum and the theoretical spectrum. In order to reduce the computational complexity, the half-flow point technique is used to estimate the initial temperature. Then the pruning polynomial fitting technique is used to obtain the minimum value and the effective temperature, and the surface gravity value and the metal abundance value are obtained in turn by the same steps. The first set of templates is used to obtain the effective temperature in the second set of templates, but the second set of templates will be used to calculate the abundance of alpha elements in future applications. In this system, we also achieve the effective temperature obtained by the observed g-r color index and the strength prediction of the Balmer (Blamer) line system. Finally, two theories of effective temperature estimation and three empirical effective temperature estimates are used. 5. The accuracy of the system parameters is evaluated by using the metallic abundance values of the globular clusters and the open clusters in the Milky way. Using the parameters extracted from high-resolution spectra observed by other telescopes as real values, the parameters of metal abundance in this system are corrected, the errors and dispersion of each algorithm in different regions are obtained, and the results are reweighted. Good accuracy has been obtained.
【學(xué)位授予單位】：山東大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2010
【分類號(hào)】：P144
【引證文獻(xiàn)】
相關(guān)碩士學(xué)位論文 前3條
1 楊琳;基于數(shù)據(jù)挖掘技術(shù)的激變變星的特征提取[D];山東大學(xué);2011年
2 韋鵬;LAMOST一維光譜自動(dòng)處理[D];山東大學(xué);2011年
3 張周周;基于Matlab和VRML的虛擬銷盤摩擦實(shí)驗(yàn)系統(tǒng)仿真設(shè)計(jì)[D];延邊大學(xué);2012年
，本文編號(hào)：2274839
本文鏈接：http://sikaile.net/kejilunwen/tianwen/2274839.html
最近更新
教材專著
