随着获取和收集天文光谱大数据能力的与日俱增, 合理利用计算科学技术正确地分析海量光谱的处理方法及结果统计。 前述工作采用了欧氏距离分析判别LAMOST实测光谱与模板之间相似度的研究, 研究恒星分类准确性取决于高质量的模板光谱, 选取LAMOST光谱在用的分类软件中183个恒星模板光谱, 分别利用欧氏距离和马氏距离方法得出A, F, G, K和M型恒星模板间的均值和最大值, 完成每条谱线相互之间的相关性分析, 找出相对距离较大的模板及形成原因。 相似度度量可视化实验数据结果表明模板之间具有一定的区分度, 通过马氏距离分析模板间相似性能更进一步辨识出相近模板之间的细微差别, 具备较优良的判别效果, 证实了LAMOST现有分类的各模板间距离较均匀, 且分类结果较为准确。 该研究可进一步优化在用光谱分类模板, 提升LAMOST恒星分类模板库的精确度和可信度。

With the vigorous development of the astronomical spectral big data acquired, such as LAMOST, assessments of the automated data reduction and analysis are necessary. The above work uses the Euclidean distance analysis to determine the similarity between LAMOST spectra and the template. The accuracy of star classification depends on the high-quality template spectra. Classification results from LAMOST 1D pipeline depend on the 183 templates, of which the dependencies should be inspected. In this paper, we calculate both Euclidean and Mahalanobis distances for each pair of templates, using these methods to get the template mean and maximum of A, F, G, K, M stars’. By completing the correlation analysis, we find that the distances averagely show similarity except for several templates. The Mahalanobis distances can even detect the difference between adjacent pairs. They can further identify that the slight differences between the similar templates have better discriminating effects. We conclude from our experiment that most of the LAMOST spectra are correctly classified, while some outstanding templates should be checked as the basis of the optimization for improving the accuracy and reliability of LAMOST templates.