Quantitative Determination Of Adulteration In Rice Based On Near Infrared Spectroscopy Analysis

The rice,as a world’s leading crop,has drawn much attention due to the quality and security issue.Recently,rice adulteration has emerged frequently,and it means that rice quality analysts have been facing unprecedented challenges.Near infrared spectroscopy(NIRS),a low cost,fast and non-destructive detecting technique,which has already applied in various fields.In this study,two common adulterations(the high grade mixed with the low grade and being polished by the mineral oil)were analyzed based on near-infrared spectroscopy and chemometrics techniques.Finally,two adulteration quantitative analysis models were established.(1)The adulteration model of high-quality rice mixed with low-quality rice was built based on NIRS combined with the content of amylase.130 high-quality rice samples were adulterated at thirteen degrees(0%,5%,10%,15%,20%,30%,40%,50%,60%,70%,80%,90% and 100%)with low-quality rice.The near infrared spectra of samples were collected,and then,the Maximum-minimum normalization,Standard Normal Variate,1st Derivative and Smoothing were used for preprocessing the spectral data,respectively.Partial least squares(PLS)model was developed for predicting the adulteration ratio.The results showed that the model using Maximum-minimum normalization pretreatment method was the best,with the coefficients(Rc,Rp)from calibration set and prediction set were 0.9698 and 0.9845 respectively,and the root mean square error of calibration(RMSEC)and root mean square error of prediction(RMSEP)were 8.66 and 6.46 respectively.(2)The adulteration model of rice mixed with mineral oil was built based on NIRS combined with partial least squares.80 rice samples were adulterated at eight degrees(0.0000%,0.0118%,0.0395%,0.1370%,0.1819%,0.6194%,0.8764% and 1.3219%)with mineral oil.And then,the near infrared spectra of samples were collected.Partial Least Squares(PLS)was applied to calibrate the prediction model for adulteration ratio.Experimental results showed the correlation coefficient and the root mean square error of the calibration set were 0.9739 and 0.106,respectively;the correlation coefficient and the root mean square error of the prediction set were 0.9888 and 0.0698,respectively.The models have very good precision and stabilization and detect two kinds of rice adulteration rapidly and accurately,providing a new perspective of rice quality control.