The Detection And Evaluation Methods Of The Eating Quality Of Rice

The Rice production of China in 2014 is more than 140 million tons,but the eating quality of Rice attracts more attention.The most direct way to evaluate the eating quality of Rice is sensory evaluation.It can feedback the complex physiological feelings brought by Rice,but has shortcomings such as high demand of experimenters,strong subjectivity,complicated operation,time consuming,instability of result and so on.With the large-scale production of Rice and Rice products,We urgently need a detection method which is stable,accurate and can classify the Rice by eating quality rapidly.In this paper,the prediction models and category discriminant models based on the electrochemical signals of single electrode sensor and sensor arrays were built for rice.The main research work and conclusions of this paper are as follows:1)Data acquisition transpose this paper used to collect rice electrochemical signals is based on square wave voltammetry and three electrode system.The reference electrode of the three electrode system is Ag/AgCl electrode,auxiliary electrode is Pt electrode and working electrodes including Au,Pd,Cu,Pb,Cd,In,Al,Ni,Fe,W,Ti,a total of 11 bare metal electrodes.After Obtaining the denoised electrochemical signals by Sym4,correlation analysis was carried out.Results show that the denoised electrochemical signals collected by Au,Pd,Cu,Pb,Cd,In,Ni,Fe,W,Ti are negative correlation with rice amylose content,positive correlation with amylopectin and the ratio of amylopectin and amylase,the denoised electrochemical signals collected by Al are negative correlation with rice amylose at some potential points and positive correlation with another potential points,amylopectin and the ratio of amylopectin and amylase are in the same condition.2)In the single electrode based rice prediction models,screening variables by variable contribution is better correlation coefficient and tolerance.Among the rice prediction models based on high-contribution variables,the best electrode to predict amylase is Al,In and Ni,R2 of these models are 0.976,0.973 and 0.898,respectively,SECV were 0.518,0.965 and 0.929.Single electrode used for amylopectin prediction is not good as amylase,the best electrode is Al,R2=0.968,SECV=2.521.Followed is Pb electrode,R2=0.895,SECV=4.367.When it comes to the ratio of amylopectin and amylase,the best electrode is Al,then In electrode,R2=0.983,0.91,SECV=0.104,0.23.In the single electrode based rice category discriminant models,the Al electrode is the most effective judging correctly 88.9%of the rice.Followed by electrode Pb,Cd,In,Ti and Cu;Ni and Au electrode is not suitable for used in classification of rice alone.3)The results of System clustering method of the 11 bare metal electrodes are as follows:(1)Al;(2)Ti;(3)Cd;(4)Cu,Au,Fe,In,Pd,Ni,W.Through the permutation and combination we can get 14 different sensor arrays.In the sensor arrays based rice prediction models,do Principal component analysis for single electrode first and do multiple linear regression second is the best choose.The amylase prediction models based on sensor arrays which composed by Al,Ti,Cd,In;Al,Ti,Cd,Ni;Al,Ti,Pb,In or A1,Ti,Pb,Fe can lead to a good results,the fit of these models are all above 0.99,up to 0.997,SECV are 0.398,0.244,0.462 and 0.5 respectively.For amylopectin,the best combination is Al,Ti,Cd,Au,R2 = 0.995,SECV= 1.248.For the ratio of amylopectin and amylase,the best combination is Al,Ti,Cd,Ni,R2 = 0.998,SECV = 0.065.The rice category discriminant models based on sensor arrays is better than the single electrode based one.Four electrode combination:Al,Ti,Cd,Cu,Al,Ti,Cd,Fe;Al,Ti,Cd,Ni and Al,Ti,Cd,W used for rice classification have higher accuracy,the accuracy of validation in the group and cross validation are 100%and 97%.4)By using the same rice samples,a series of prediction models and category discriminant models were built for rice based on near infrared diffuse reflection.Results follows:Normalized conbined with PCR is the best one which can eliminate the tiny optical path difference and extract the concentration signal.It has three principal components PCI,PC2,PC3 for amylase model,amylopectin model and their radio model,the fit of the prediction models are 0.882,0.697,0.763;SECV are 0.351,5.775,0.259.The accuracy of validation in the group and cross validation of rice discriminant model are both 75%.5)Compared with near infrared spectroscopy,electrochemical technique is more accurate.It is thanks to more scanning point the electrochemical technology use,makes the sample information more comprehensive.