摘要
基于径向基函数神经网络(RBFN)建立了茶多酚总儿茶素含量的近红外光谱分析模型。茶多酚光谱采用小波压缩、标准化处理后,进行主成分分解,以主成分光谱作为RBFN的隐层输入函数,并通过改变主成分数对模型进行优化。当主成分数为7时得到了RBFN优化模型,该模型对定标样品集、全样品集和预测样品集的预测值与实际值回归系数R分别为0.999,0.999和0.992,预测均方误RMSEP分别为1.08%,2.06%和3.68%。
A near infrared spectroscopic models for determinating the total catechins in tea polyphenol powder were presented based on radical basis function network. The spectra of tea polyphenol powder samples were pretreated with wavelet transform (sym6), standard normalization and principle components analysis (PCA), the eigenvectors of PCA were used as the input functions of RBFN. The models were optimized by changing the number of PCA and evaluated by the correlation coefficients (R) and root mean square error of prediction (RMSEP) of calibration set and the prediction set. The optimum correlation coefficient (R) of the calibration set, all sample set and prediction set were 0. 999, 0. 999 and 0. 992, respectively, and RMSEP of them were 1.08%, 2.06% and 3.68%, respectively.
出处
《光学精密工程》
EI
CAS
CSCD
北大核心
2006年第1期58-62,共5页
Optics and Precision Engineering
基金
吉林省科技计划重点项目(No.20040324-2)
中国科学院长春光学精密机械与物理研究所应用光学国家重点实验室开放基金
浙江省自然科学基金项目(No.202081)
关键词
径向基函数网络
茶多酚
儿茶素
近红外光谱
模型
radical basis function network (RBFN)
tea polyphenol
catechins, near infrared spectroscopy (NIR)
model