基于多元散射校正和偏最小二乘(MSC/PLS)的傅里叶变换近红外光谱检测蜂蜜中还原糖

Near-infrared determination of reducing sugar content in honey based on multiplicative scatter correction partial least square (MSC/PLS) method

为快速无损测定蜂蜜的还原糖,提出一种用近红外光谱技术进行检测的新方法。选用SN/T0852——2000铁氰化钾滴定法作为光谱还原糖测定的标定方法,用不同光谱预处理方法对模型的影响进行比较。首先用MSC对光谱数据进行校正,在经过Norris导数平滑法对原始光谱进行预处理消除噪声。利用光谱专用分析软件TQAnalyst7.2,建立偏最小二乘(PLS)校正模型。最优光谱模型校正相关系数(R)、校正相关均方差(RMSEC)、预测相关均方差(RMSEP)、交互验证相关系数(RCV)、交叉验证均方差(REMSECV)分别为0.97056,1.21,1.54,0.96267,1.36。结果表明,应用多元散射校正(MSC)处理,并经过Norris二阶导数平滑后的PLS模型优于经典PLS算法所建立的模型。

A new method of detection for rapid reducing sugar deter rnination of honey without destroying was developed. Reducing sugar calibration method was potassium ferrocyanide titration in SN/ T0852 -2000. Different pretreatment methods for spectrum model were compared. Multiplicative scatter correction was used as a method to preprocess the near infrared（ NlR） spectra of honey ranging from4 000-10000 cm^-1. Eslahlishing the calibration model of reducing sugar content against honey spectra before and after MSC pretreatment by partial least square（PLS）. The optimal spectrum model correction correlation coefficient （R）, Root mean standard error of calibration （RMSEC）, Root mean standard error of prediction （RMSEP）, Root mean interaction validation correlation coefficient （Roy） Root mean standard error of cross validation （REMSECV） respectively are 0. 970 56, 1.21, 1.54, 0. 962 67, 1.36.