Study On Wheat Protein Content By Near-infrared Non-destructive Testing Method

The protein content of wheat is one of the important criteria for evaluating the quality of wheat.So how to test the protein content in wheat fast,effectively,and non-destructively is one of the studies all over the world.When the measurement of wheat by near infrared spectroscopy,spectral information collected there is a lot of noise,but also accompanied by on information and interference variables.In order to solve the problem above,this thesis intends to research the non-destructive testing method of protein content in wheat.The main research contents are as follows:(1)In order to eliminate the influence of external noise,the author of this thesis utilizes wavelet packet(WTP)to make a noise reduction of the original spectral data.(2)In order to delete the impact of uninformative variables and disturbing variables,the author of this thesis combines two kinds of information vectors(IVs)weighted model with model population analysis(MPA)for the first time and put forward the automatic weighting variable combination population analysis(AWVCPA).In addition,the author of this thesis analyzes the impact of sample set variation on the choosing method of every variation.The monte carlo variable combination population analysis(MC-VCPA)is put forward by the combination of monte carlo sampling techniques(MCS)and variable combination population analysis(VCPA).(3)Utilize partial least squares(PLS)to establish the prediction model of protein in wheat.Compared with other modeling methods,the results show that the variables of WTP-AWCVPA-PLS are 9,and the variables of WTP-MC-VCPA-PLS are 10.Predicted root mean square error(RMSEP)decreases from 0.5096 to 0.3382 and 0.3295 compared with the original spectral modeling.Respectively increase 34% and 35%.In comparison with other modeling methods,theses two methods not only utilize the least variables,but also has the highest accuracy.