基于非参数回归与Adaboost的恒星光谱自动分类方法

发布时间：2018-04-17 08:44

本文选题：光谱分类 + Adaboost　；参考：《光谱学与光谱分析》2017年05期

【摘要】：通过对恒星光谱进行分析可以研究银河系的演化与结构等科学问题,光谱分类是恒星光谱分析的基本任务之一。提出了一种结合非参数回归与Adaboost对恒星光谱进行MK分类的方法,将恒星按光谱型和光度型进行分类,并识别其光谱型的次型。恒星光谱的光谱型及其次型代表了恒星的表面有效温度,而光度型则代表了恒星的发光强度。在同一种光谱型下,光度型反映了谱线形状细节的变化,因此光度型的分类必须在光谱型分类基础上进行。本文把光谱型的分类问题转化为对类别的回归问题,采用非参数回归方法进行恒星光谱型和光谱次型的分类;基于Adaboost方法组合一组K近邻分类器进行光度型分类,Adaboost将一组弱分类器加权组合产生一个强分类器,提升光度型的识别率。实验验证了所提出分类方法的有效性,光谱次型识别的精度达到0.22,光度型的分类正确率达到84%以上。实验还对比了两种KNN方法与Adaboost方法的光度型分类,结果表明,利用KNN方法对光度型分类精度低,而基于弱分类器KNN的Adaboost方法将识别率大幅提升。
[Abstract]:The evolution and structure of the Milky way Galaxy can be studied by analyzing the spectra of stars. Spectral classification is the basic Ren Wuzhi in the spectral analysis of stars.In this paper, a method of combining non-parametric regression with Adaboost to classify the spectrum of stars is proposed. The stars are classified according to spectral type and photometric type, and the subtypes of their spectral types are identified.The spectral type and its subtype of the star spectrum represent the surface effective temperature of the star, while the luminosity type represents the luminous intensity of the star.Under the same spectral pattern, the photometric type reflects the variation of the shape details of the spectral line, so the classification of the photometric type must be based on the spectral type classification.In this paper, the classification problem of spectral type is transformed into the regression problem of category, and the classification of spectral type and spectral subtype of star is carried out by non-parametric regression method.Based on the Adaboost method, a group of K-nearest neighbor classifiers are combined for photometric classification Adaboost. A group of weak classifiers are weighted to produce a strong classifier to improve the recognition rate of photometric type.The experimental results show that the proposed classification method is effective. The accuracy of spectral subtype recognition is 0.22, and the accuracy of photometric classification is over 84%.The experiment also compares the photometric classification of two KNN and Adaboost methods. The results show that the accuracy of the photometric classification is low by using the KNN method, while the recognition rate is greatly improved by the Adaboost method based on the weak classifier KNN.
【作者单位】：北京服装学院基础部;西安建筑科技大学理学院;中国科学院国家天文台;北京师范大学信息科学与技术学院;
【基金】：国家自然科学基金委员会-中国科学院天文联合基金项目(U1531242)资助
【分类号】：P144.1

【相似文献】