在线检测无花果中可溶性固形物的近红外漫透射技术研究

摘要

采用近红外光谱漫透射技术,建立无花果可溶性固形物(SSC)的在线、无损、快速检测方法,对于无花果品质的快速分级具有重要意义.试验对203个无花果进行了近红外光谱采集与SSC测量,采用偏最小二乘(PLS)建立SSC的预测模型,比较了不同一阶导数窗口宽度对模型结果的影响,确定最佳宽度为61,并采用无信息变量消除法(UVE)进行波长优选.结果表明,UVE-PLS能够有效简化模型,波长变量由1010个降低到211个,同时PLS模型的精度也得到了很大的提高,校正均方根误差(RMSEC)、交叉验证均方根误差(RMSECV)、预测均方根误差(RMSEP)分别为0.63 °Brix、0.78°Brix、0.83 ° Brix,校正相关系数(RC)、交叉验证相关系数(RCV)、预测相关系数(RP)分别为0.89、0.83、0.83.本研究为无花果品质的在线无损检测提供了有效的理论依据与实践经验.%It is of great importance for the fast quality classification of figs usingthe fast, nondestructive and on-line determination ofsoluble solids content(SSC)based on near infrared(NIR)spectroscopy.The diffuse trans-missionNIR spectra were obtained from a total of 203 figs by aQE65 prospectrometer and the SSC in figs was measured using a portable refractometer.Thenthe partial least squares(PLS)was used to build the prediction model for SSC.After compared the effect of the spectral variables width of 1st derivative on the model accuracy, the optimal width 61is obtained.Uninformative variable eliminationPLS(UVE)was used for wavelength selec-tion.The results show that UVE simplifies the model,and the wavelength variables are reduced from 1010 to 211 and the precision of theUVE-PLS model is enhanced.The root mean square error of calculation(RMSEC),root mean square error of cross-validation(RMSECV)and root mean square error of prediction(RMSEP)are 0.63° Brix,0.78°Brix and0.83°Brix,respectively,and the correlation coefficient of calibration(RC),correlation co-efficient of cross-validation(RCV)and correlation coefficient of prediction(RP)are 0.89,0.83 and 0.83,re-spectively.This work supplies effective theory and practical experience for the un-destructive on-line quality de-tection of figs.