基于红外与近红外光谱的烟叶部位识别  

Recognition of Plant Parts of Tobaccos Based on Infrared and Near Infrared Spectra

作　　者：张鑫[1] 郭佳[1] 倪力军[1] 张立国[1] 

机构地区：[1]华东理工大学化学与分子工程学院,上海200237

出　　处：《光谱学与光谱分析》2007年第12期2437-2440,共4页Spectroscopy and Spectral Analysis

基　　金：科技部"十五"攻关项目(2001BA701A04)资助

摘　　要：以烟叶样品的红外及近红外光谱为基础,采用基于主成分分析的马氏距离判别模型,研究了不同类型仪器、建模区间、模型参数及光谱预处理方式对烟叶部位识别准确率的影响。结果表明根据红外和近红外光谱均可对烟叶部位进行良好识别,近红外光谱因包含的样品信息更为丰富,可以得到比红外光谱更好的识别效果。其中仪器A的二阶导数光谱给出的烟叶部位识别准确率最高,可达94.11%;仪器B的一阶导数及SNV光谱给出的烟叶部位识别准确率次之,为88.24%;Nicolet公司的Antaris360傅里叶红外仪的一阶导数光谱给出的烟叶部位识别准确率为82.35%。对于同一仪器,最佳建模区间及主成分个数随样本情况及光谱预处理方式而变。In the present paper, an IR/NIR spectrometry based the principal component analysis （PCA） space was established on the pattern recognition using Mahalanobis distance method in for the discrimination of plant parts of tobaccos. Effects of the type of IR/NIR spectrometers, calibration region of the spectra, model parameters and pretreatment of the spectra on the accuracy of discrimination were investigated using tobaccos cultivated in Yunan Province in 2003 and 2005 as case study. The recognition model shows the internal relationships between the information of spectra and the plant parts of tobaccos. The results indicate that both IR and NIR eould be successfully used to recognize plant parts of the tobaccos, but the latter was better because it involves more sample information. It was found that the highest recognition accuracy, 94. 11%, was obtained by using apparatus A with the second derivative spectra, while recognition accuracy of 88.24% and 82.35% was respectively given by apparatus B with with the first derivative SNV spectra and IR spectrometer with first derivative spectra. For the same spectrometer, the optimal calibration region and principal component number were changed with samples and the spectral pretreatment method.

关 键 词：红外 近红外 烟叶部位 模式识别 

分 类 号：O657.3[理学—分析化学]