可见／近红外光谱预测杨梅汁酸度的方法研究

针对可见/近红外光与杨梅汁酸度存在非线性相关的特点,提出了应用偏最小二乘(PLS)法预测线性部分和人工神经网络(ANN)预测非线性部分,结合两种方法综合预测杨梅汁酸度值,通过比较,r RMSEP,B ias的值来检验该方法.其中PLS模型用于寻找与杨梅汁酸度值有关的敏感波段,预测杨梅汁酸度的线性部分,将这些敏感波段对应的光谱吸光度值作为人工神经网络的输入,并将杨梅汁酸度的实际测量值减去PLS模型校正值,获得的差额部分作为神经网络的输出,建立一个差额神经网络预测杨梅汁酸度的非线性部分.46个样本用于建模,30个样本用于预测.结果表明该方法对样本的预测相关系数r=0.939,RMSEP=0.218,B ias=-0.121,好于只使用PLS模型的相关系数r=0.921,RMSEP=0.228,B ias=-0.132.

METHOD FOR PREDICTING ACIDITY OF BAYBERRY JUICE BY USING VIS/NEAR INFRARED SPECTRA

Aiming at the nonlinear correlation characteristic of visible/near infrared spectra and the corresponding acidity of bayberry juice,one mixed algorithm was presented to predict the acidity of bayberry juice with partial least squares(PLS) and artificial neural network(ANN).The values of correlation coefficient(r),the root mean squared error of prediction(RMSEP),and bias were used to estimate the mixed model.PLS was used to find some sensitive spectra related to acidity in juice,and the values of spectral absorptance corresponding to them were regarded as the input neurons of ANN.Remnant values by subtracting standard values and validation values were regarded as the output neurons of ANN.The calibration equation developed from them was used to predict the constituent values for the independent spectra of 30 samples.The results indicate that the observed results by using PLS-ANN(r=0.939,RMSEP=0.218,Bias=-0.121) are better than those obtained by PLS(r=0.921,RMSEP=0.228,Bias=-0.132).