摘要
提出一类红外光谱分析数据特征指纹的可视化表达方法。该方法采用基于贝叶斯准则的多元统计方法对原始数据进行投影变换 ,再以二维灰度图对变换后数据进行可视化表征 ,形成基于计算机图像的虚拟化学指纹图谱。将其用于鉴别 3种不同产地中药当归样品的结果表明 ,它能有效提取红外光谱分析数据的特征指纹 ,实现以虚拟指纹图谱对药材产地的分类鉴别 。
A new method for representing fingerprint feature by visualization of infrared spectral data was proposed. Using the multivariate statistical method based on Bayes rules, raw analytical data were transformed into a new subspace by mathematical projection. Subsequently, the transformed data were visualized into the virtual chemical fingerprint plots by means of two-dimensional grayscale images. The proposed method was applied to identify herbal medicine-DangGui from three origins. The results showed that it can effectively extract the image characteristics from infrared spectral data, and the different geographical origins of DangGui samples can be accurately identified by the virtual chemical fingerprint plots. Therefore, it provides a new technical tool for identifying complicated chemical substance system.
出处
《分析化学》
SCIE
EI
CAS
CSCD
北大核心
2002年第12期1426-1430,共5页
Chinese Journal of Analytical Chemistry
基金
国家重点基础研究发展规划 ( 973计划
No .G19990 5 440 5 )
国家自然科学基金 (No .39870 940 )资助项目
