基于中红外漫反射光谱技术测定精米中直链淀粉含量的研究

Amylose Content Detection in Polished Rice Based on Diffuse Reflectance Fourier Transform Infrared Spectroscopy

研究了基于傅里叶中红外漫反射光谱(DR-FTIR)技术结合偏最小二乘法(PLS)、主成分分析法(PCR)、经典最小二乘法(CLS)、逐步多元线性回归法(SMLR)快速测定精米中直链淀粉含量。实验选取了161个精米样品作为研究对象,通过光谱预处理,异常值剔除,波谱区间选择对模型进行了优化。结果表明:采用PLS建模,通过多元散射校正,向后区间偏最小二乘法预处理后,所构建的模型效果最优。此时,建模区间为3 800~3 500、3 100~3 000、2 900~2 400、2 300~1 300、1 000~900 cm^(-1),模型相关系数R为0.995 6,校正均方差为0.291,预测均方差为1.23;同时,实验另选了34种精米样品对模型进行了准确性验证。结果显示验证集样品的红外预测值与真实值高度相关,线性拟合方程为Y=0.994X+0.068,相关系数R为0.994 9,说明实验所开发的直链淀粉DR-FTIR模型是有效的,预测结果准确度好,模型稳定性高。

The analysis method for rapidly determining amylose content in milled rice by Diffuse Reflectance Fourier transform infrared spectroscopy（ DR-FTIR） combined with partial least squares（ PLS）,principal component regression（ PCR）,classical least squres（ CLS）,stepwise multiple linear regression（ SMLR） were studied. With a set of 161 milled rice samples as study objects in this research,the model was optimized through spectra pretreatment,abnormal value removing and spectra region selecting. The result showed that the optimal model for the analysis of amylose content was obtained through using PLS,multiplicatiove signal correction（ MSC）,and back interval partial least squares（ bi PLS）. Under these conditions,the optimal modeling intervals were the wavenumber in the range of3 800 ～ 3 500,3 100 ～ 3 000,2 900 ～ 2 400,2 300 ～ 1 300 and 1 000 ～ 900 cm^-1,the correlation coefficient R（0. 995 6）,root mean squreerror of calibration（RMSEC） of 0. 291,root mean squreerror of prediction（RMSEP） of1. 23. Meanwhile the research selected the other 34 milled rice samples to verify the model accuracy. The result showed that the FTIR predictive value were highly correlated with the actual values with a regression equation of Y =0. 994 X ＋ 0. 068,the correlation coefficient R（0. 994 9）,which showed that the amylose content model of milled rice developed by DR-FTIR was effective with accurate prediction result and high model stablity.