NIRS method is used to build chrysanthemum total flavone in quick content determination method study and the model establishment method and spectral preprocessing method and model best latent variable number were screened determine the PLS model is established, with SNV method combined with 1D, nd method to preprocess the original chrysanthemum NIR spectra, and by comparing the predicted residual sum of squares and (press) and cross validation were root mean square error (RMSECV) and other parameters to determine the model of optimal number of latent variables is 8。 Use of these methods to establish the calibration model, the correlation coefficient is 0。8329 correction error RMS (RMSEC) is 0。0661 predicted root mean square error of prediction (RMSEP) 0。1000, predictive value and true value of good correlation, indicating a PLS model can effectively predict the content of total flavonoids in Chrysanthemum samples of unknown content。 This method will study the rapid quality evaluation system on Chrysanthemum provided scientific basis。

毕业论文关键词：菊花； 近红外光谱； 黄酮； 含量测定； 质量评价

Keyword: Chrysanthemum; NIRS; flavone; determination of content; quality evaluation

缩略词表

1D first derivative 一阶导数

2D second derivative 二阶导数

A absorbance 吸光度

C content 含量

S original spectra value 原始光谱值

n sample number 样本数

NIR near infrared spectra 近红外光谱

NIRS near infrared spectroscopy 近红外光谱法

OS original spectra 原始光谱

PCR principle component regression 主成分回归

PLS partial least square 偏最小二乘

R correlation coefficient 相关系数

RMSECV root mean square error of cross-validation 交叉验证均方根误差

PRESS predicted residual sum of squares 预测残差平方和

SMLR stepwise multiple linear regression 逐步多元线性回归

RSD relative standard deviation 相对标准偏差

RMSEC root mean square error of calibration 校正集均方根误差

SG Savitzky-Golay filter smoothing SG平滑

ND Norris derivative filter smoothing Norris 导数平滑

Rp correlation coefficient of prediction 预测集相关系数

RMSEP root mean square error of prediction

预测均方根误差

MSC