恒星光谱数据弱特征识别方法

Method for Recognizing Weak Features of Stellar Spectral Data

恒星光谱弱特征识别是LAMOST光谱数据分析的重要研究内容,能够为恒星光谱分类提供重要科学依据。目前,针对恒星光谱数据进行特征识别的方法较多,但是缺乏对某种特定特征谱线进行精确提取的算法。针对LAMOST低分辨光谱数据中Hα弱发射线轮廓形态多样问题,提出了一种基于置信度的Hα弱发射线识别方法。首先,基于Hα弱发射线轮廓形态特征给出Hα弱发射线的置信度的度量方法。利用Hα发射线波长区间内峰值与发射线的偏移量建立距离置信度模型,根据高斯轮廓所含像素点个数建立高斯轮廓副信息模型,通过计算峰值左右波形的差异建立对称性评估模型,结合三个模型给出最终的Hα弱发射线的置信度,并基于此置信度进行第一轮筛选。为了提高精度,提出了借助其它发射线的特征给出了基于二分类的Hα发射线筛选策略。通过考察Hβ、NII、OIII以及SII发射线的特征,基于辅助信息的决策树进行第二轮筛选,进一步提高筛选的精度。实验结果表明:提出的Hα弱发射线的特征度量方法的准确度高达90%,并且速度较快,平均每1 k数据耗时仅三十多秒。

The massive spectral data in LAMOST provides precious samples for scientific research in many fields including astronomy.It is an important research to identify weak features of stellar spectra for spectral data analysis,which can provide an important scientific basis for stellar spectral classification.At present,there are many methods for feature recognition based on stellar spectrum data,but few of them can accurately extract certain feature lines.Aiming at the diversified profile of Hαweak emission lines in LAMOST low-resolution spectral data,a method for identifying Hαweak emission lines based on confidence is proposed.First,based on the profile characteristics of the Hαweak emission line,a measure of the confidence of the Hαweak emission line is given.The distance confidence model is established by using the offset between the peak value and the emission line in the wavelength range of the Hαemission line,the Gaussian contour side information model is established according to the number of pixels contained in the Gaussian contour,and the symmetry evaluation model is established by calculating the difference between the waveforms on the left and right sides of the peak.The three models are combined to give the confidence of the final Hαweak emission line,and the first round of screening is performed based on this confidence.In order to improve the accuracy,it is proposed to use the characteristics of other emission lines to give a Hαemission line screening strategy based on two classifications.By examining the characteristics of Hβ,NII,OIII and SII emission lines,the decision tree based on auxiliary information is used for the second round of screening to further improve the accuracy of screening.Experimental results show that the accuracy of the proposed Hαweak emission line feature measurement method is as high as 90%,and the speed is relatively fast,with an average of only more than 30 seconds per 1 k data.