基于拉曼光谱的广陈皮年份检测方法  

Age Detection Method of Citrus reticulata Based on Raman Spectroscopy

作　　者：李静敏 辛志昂 聂青青 罗甲 白华[1] 高强[1] LI Jingmin;XIN Zhiang;NIE Qingqing;LUO Jia;BAI Hua;GAO Qiang(Tiangong University,School of Electronic&Information Engineering,Tianjin 300387)

机构地区：[1]天津工业大学电子与信息工程学院,天津300387

出　　处：《食品工业》2023年第1期279-285,共7页The Food Industry

基　　金：2021年国家级大学生创新创业训练计划项目“基于光谱指纹特征和模式识别的广陈皮年份检测方法研究”(No.202110058027)。

摘　　要：分别采集不同年份广陈皮正面和反面的拉曼光谱数据,利用光谱预处理方法结合t检验的统计学分析方法对广陈皮进行定性和定量的分析,并且利用主成分分析的模式识别法构建广陈皮年份鉴别模型。根据t检验的结果发现,广陈皮反面的拉曼特征峰大部分均通过t检验(P<0.05),说明反面的拉曼光谱在不同年份间具有显著性差异,同时反面768, 869, 1 598和2 124 cm^(-1)处特征峰随年份的单调变化,说明可以利用这4个特征峰鉴别广陈皮的年份。将预处理后的81组反面光谱的4个特征峰数据作为训练集,取均值后新生成的39组反面数据作为测试集,通过主成分回归法建立广陈皮年份鉴别预测模型,训练集预测精度为77.78%,测试集预测精度为76.92%,初步实现不同年份广陈皮的鉴别预测。In this paper, Raman spectrum data of the front and back sides of Citrus reticulata in different years were collected,the qualitative and quantitative analysis of Citrus reticulata was carried out by using spectral preprocessing method combined with statistical analysis method of student’s t-test, and the year identification model of Citrus reticulata was constructed by using pattern recognition method of principal component analysis. According to the results of student’s t-test, it was found that most of the Raman characteristic peaks on the reverse side of Citrus reticulata passed the student’s t-test(P<0.05), indicating that the Raman spectroscopy on the reverse side of Citrus reticulata had significant differences among different years. At the same time, the Raman characteristic peaks at 768, 869, 1 598, 2 124 cm^(-1) on the reverse side changed monotonously with the year, indicating that these four characteristic peaks could be used to identify the year of Citrus reticulata. The four characteristic peak data of 81 groups of reverse data after pretreatment were taken as the training set, and the 39 groups of reverse data newly generated after taking the mean value were taken as the test set. Through the principal component regression method, the year identification prediction model of Citrus reticulata was established. The prediction accuracy of the training set was 77.78%, and the prediction accuracy of the test set was 76.92%. The identification prediction of Citrus reticulata in different years was preliminarily realized.

关 键 词：拉曼光谱 主成分分析 T检验 广陈皮 模式识别 

分 类 号：R284.1[医药卫生—中药学] O657.37[医药卫生—中医学]