Research On Rice Quality Detection Based On Terahertz Time Domain Spectroscopy

The issue of food security is related to the stability of the country and the guarantee of the quality of life of the people.It is one of the issues that the people are most concerned about.Terahertz spectroscopy technology,as an emerging spectrum detection technology,has its unique advantages in food safety detection.In this paper,using terahertz spectroscopy combined with chemometrics methods,with rice,rice and colored rice as the research object,comprehensive testing of rice quality.The main research results and conclusions are as follows:First of all,rice with germination,mildew and different degrees of mildew is studied.Four terahertz time domain spectroscopy and chemometrics methods are used to classify and identify the four samples.Use PLS-DA,principal component analysis and LS-SVM to predict and model the data.Using the data after the second-order derivative preprocessing method for principal component analysis and PLS-DA predictive modeling can accurately distinguish the four qualities of rice.The LS-SVM modeling prediction results using raw data and RBF kernel functions are the best,and the prediction accuracy can reach 100%.Based on the research of moldy rice,rice with different degrees of mildew are taken as the research object,and the terahertz technology is used to distinguish rice with 2,4,6,8 and 10 days of mildew.The results of using PLS-DA and principal component analysis are not ideal.The modeling and prediction effect using LS-SVM is very stable.The prediction accuracy of LS-SVM using Lin kernel function and RBF kernel function can reach 98%.Then,the three long-grained rice varieties are selected as the research object.Extract the first three principal components of all spectral absorption data for three-dimensional clustering analysis of principal components.In the distribution of the three-dimensional map,Jiangnian Wannian Gongmi and Northeast Long Grain Fragrant Rice have partial overlaps;PLS-DA modeling prediction,The false positive rate is 3.3%.The effect of modeling prediction using LS-SVM is still the best,and the prediction accuracy of LS-SVM using RBF kernel function is 100%.Finally,the colored rice(purple rice and black rice)is taken as the research object to quantitatively detect the adulteration of colored rice.Firstly,the three samples of rice,purple rice mixed and dyed rice and purple rice mixed and dyed black rice are identified by simple qualitative classification.Quantitative detection of adulterated samples with different quality scores using PLS and LS-SVM,respectively.The data after baseline correction preprocessing combined with least squares support vector machine(LS-SVM)is the best method for quantitative modeling.Among them,purple The predicted correlation coefficient(Rp)of rice mixed with dyed rice is 0.979,and the predicted root mean square error(RMSEP)is 0.091;the predicted correlation coefficient(Rp)of purple rice mixed with dyed black rice is 0.948,and the predicted root mean square error(RMSEP)is 0.093.This paper uses terahertz spectroscopy combined with LS-SVM algorithm modeling to accurately detect rice quality,and provides a certain reference and basis for the detection of terahertz time-domain spectroscopy in food quality and safety.