基于特征约束的谱线自动提取  

Auto-Extraction of Spectral Lines Based on Feature Constraints

作　　者：刘蓉[1] 段福庆[2] 刘三阳[1] 

机构地区：[1]西安电子科技大学数学系,陕西西安710071 [2]北京师范大学信息科学与技术学院,北京100875

出　　处：《光谱学与光谱分析》2007年第8期1648-1652,共5页Spectroscopy and Spectral Analysis

基　　金："863"计划项目(2003AA133060)资助

摘　　要：已有的谱线自动提取方法均采用整体阈值约束或局部阈值约束进行谱线识别,因而谱线提取结果中普遍存在谱线遗失或伪谱线过多的缺点。文章在谱线识别时加入了2个特征约束,第1个特征约束是:谱线线心的强度必须大于局部阈值和整体下阈值,并且如果某一点的强度大于整体上阈值,则可认为在该点存在谱线;第2个特征约束是:谱线的起始波长和终止波长处的强度必须小于谱线线心的强度。这2个特征约束使得该文的谱线提取效果较之已有方法有显著的提高。通过实验对该文方法和已有的方法进行了比较,实验结果充分体现了该文方法的优势。By using single thresholding or local thresholding in spectral line recognition, nearly all methods for spectral line autoextraction have the defect that there are many pseudo spectral lines or some spectral lines are lost. The present paper uses two feature constraints in spectral line recognition. The first constraint is that the central intensity of a spectral line must be higher than both the lower global threshold and the local threshold, and the point where the intensity is higher than both the upper global threshold must be on a certain spectral line. The second one is that the intensities at initial position and end position of a spectral line must be lower than its central intensity. The two feature constraints play a key role in improving the quality of spectral line extraction. Experiments show that this method is superior to the techniques used in the literature.

关 键 词：谱线提取 天体光谱 特征约束 小波变换 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]