一种新的恒星光谱间距离度量方法:残差分布距离

发布时间：2018-04-24 08:52

本文选题：恒星光谱 + 距离度量　；参考：《光谱学与光谱分析》2015年12期

【摘要】：距离度量是光谱巡天数据处理中的一个重要研究内容,其定义了一种不同光谱间的距离计算方法,以此为基础可进行光谱的分类、聚类、参数测量及离群数据挖掘等工作。距离度量方法的好坏在一定程度上影响了分类、聚类、参数测量及离群数据挖掘的效果及性能,同时随着大规模恒星光谱巡天项目的开展,如何针对恒星光谱定义更为有效的距离度量方法成为其数据处理中一个非常关键的问题。基于此问题,在充分考虑到恒星光谱的特点及其数据特征的基础上,提出一种新的恒星光谱间的距离度量方法:残差分布距离。该距离度量有别于传统计算恒星光谱间距离计算方法,利用该方法计算恒星光谱间的距离时,首先将两条光谱归一化到同一尺度下,然后计算对应波长处的残差,以残差谱分布的标准差作为距离度量。该距离度量方法可用于恒星分类、聚类以及恒星大气物理参数测量等应用中。本文以恒星光谱细分类为例来比较检验该距离度量方法,结果表明该方法定义的距离在分类时能更为有效的刻画不同类别光谱间的差距,可以很好的用于相关应用中。同时还研究了信噪比对该距离度量方法的影响:残差分布距离一定程度上受光谱信噪比影响,信噪比越小,对距离的影响越大;在信噪比大于10之后,残差分布距离对分类的影响很小。
[Abstract]:Distance measurement is an important research content in spectral survey data processing. It defines a distance calculation method between different spectra, which can be used for spectral classification, clustering, parameter measurement and outlier data mining. To some extent, distance measurement affects the effect and performance of classification, clustering, parameter measurement and outlier data mining. At the same time, with the development of large-scale star spectral survey project, How to define a more effective distance measurement method for star spectrum is a key problem in data processing. Based on this problem and considering the characteristics of star spectrum and its data characteristics, a new distance measurement method of star spectrum is proposed: the distance of residual distribution. The distance measure is different from the traditional method of calculating the distance between the spectra of stars. When the distance between the spectra of stars is calculated by this method, the two spectra are normalized to the same scale, and then the residuals at the corresponding wavelengths are calculated. The standard deviation of the residual spectrum distribution is taken as the distance measure. The distance measurement method can be used in star classification, clustering and measurement of atmospheric physical parameters of stars. In this paper, the star spectral fine classification is taken as an example to compare and test the distance measurement method. The results show that the distance defined by this method can more effectively depict the difference between different categories of spectra in classification, and can be used in related applications. At the same time, the influence of SNR on the distance measurement is studied: the residual distribution distance is affected by the spectral SNR to some extent, the smaller the SNR is, the greater the influence on the distance is, and after the SNR is greater than 10, The distance of residual distribution has little effect on classification.
【作者单位】：山东大学(威海)机电与信息工程学院;中国科学院光学天文重点实验室中国科学院国家天文台;
【基金】：国家自然科学基金项目(U1431102,11473019,11303036)资助
【分类号】：P144.1

【相似文献】