Research On Transgenic,origin Identification And Protein Etc.quantitative Detection Methods Of Soybeans Based On Terahertz Spectroscopy

Soybean is the most important oil crop and high protein feed material in China,and it is also an agricultural product that my country needs to import in large quantities.Given that China is already adopting genetically modified soybeans as processing ingredients for food and other products,the safety management of genetically modified soybeans needs to be continuously strengthened.Soybean traceability supervision is also an inevitable trend of future development.At the same time,traditional qualitative identification methods and quantitative detection methods of protein and other components often have problems such as high cost,low efficiency,and complicated operation.Therefore,it is necessary to study a method for transgenic,origin identification and protein etc.quantitative detection of soybeans.Compared with traditional qualitative identification methods and protein etc.quantitative detection methods of soybeans,Terahertz(THz)spectroscopy technology that has emerged in recent years has its unique advantages.The vibration,rotation of molecules and the interactions between molecules such as hydrogen bonds have many characteristic absorption peaks in the THz frequency band.These absorption characteristics are the unique fingerprint absorption spectrum of these substances.Thus,THz spectroscopy technology is very sensitive to detect small differences and changes in the structure of matter.It can be used for the identification and quantitative determination of these substances.Therefore,it is necessary to study the rapid and accurate qualitative identification method based on THz spectroscopy for issues such as genetically modified and origin.At the same time,how to achieve the quantitative detection of protein content,acid value,peroxide value and vitamin E in soybeans by THz spectroscopy is still a problem very worthy of study.At present,the research on the THz absorption spectroscopy of different chemical components is still in the initial stage,and the understanding of the absorption peak positions of different chemical components in the THz frequency band is also very limited.From a microscopic point of view,how to simulate the absorption peak position of vitamin E in soybeans through density functional theory(DFT)and THz absorption spectra.Moreover,the selection of characteristic spectral regions for quantitative detection of vitamin E in soybeans based on the positions of absorption peaks is still in an exploratory stage.In order to monitor the import of genetically modified soybeans,determine the origin of soybeans,and improve protein etc.quantitative detection methods.Therefore,research on transgenic,origin identification and protein etc.quantitative detection methods of soybeans based on terahertz spectroscopy has become the most urgent task nowadays.This thesis was mainly based on THz spectroscopy to research the soybean genetically modified,origin identification and protein etc.quantitative detection methods.This thesis was mainly divided into three major parts,namely: Research on soybean genetically modified and origin identification based on THz frequency-domain spectroscopy;Research on methods for quantitative detection of protein,acid value and peroxide value in soybean based on THz absorption spectroscopy;Research on the selection method of characteristic spectrum for the quantitative detection of vitamin E in soybean based on DFT and THz absorption spectroscopy.The main research contents and conclusions of this paper were as follows:(1)Research on the identification model and optimization algorithm of transgenic and non-transgenic soybeans based on THz frequency-domain spectroscopy.THz frequency-domain spectroscopy was used to identify transgenic and non-transgenic soybean samples.The THz frequency-domain spectral characteristics and differences between transgenic and non-transgenic samples were analyzed.The qualitative identification model for transgenic and non-transgenic soybeans(referred to as transgenic identification model)based on THz frequency-domain spectroscopy was established.After that,the method of transgenic and non-transgenic soybeans based on the THz frequency-domain spectroscopy and spectral region optimization algorithm was investigated.The experimental results showed that THz frequency-domain spectroscopy could be used to identification transgenic and non-transgenic soybeans.Compared with discriminant partial least squares(DPLS)and particle swarm optimization-support vector machine(PSO-SVM)transgenic identification models,grey wolf optimizationsupport vector machine(GWO-SVM)transgenic identification model combined with the first derivative pre-processing could obtain better verification results.The total identification accuracy was 96.49%(100% for transgenic and 93.55% for nontransgenic),and the identification time was 33.87 s.This indicated that this method was a way to rapidly distinguish transgenic and non-transgenic soybeans.The spectral range optimization algorithm improved the total accuracy and speed of identifying transgenic and non-transgenic soybeans by THz frequency-domain spectroscopy.After the optimizing of spectral range and mean centering pre-processing,the total identification accuracy of grid search-support vector machine(Grid search-SVM)transgenic identification model was 98.25%(96.15% for transgenic and 100% for non-transgenic).This indicated that this method was a fast and accurate means to identifying transgenic and non-transgenic soybeans.(2)Research on soybean origin identification and optimization algorithm based on THz frequency-domain spectroscopy.The differences in THz frequency-domain spectra of experimental samples from different origins were studied.The THz frequencydomain spectrum of the sample were analyzed,and the optimal modeling spectral range was selected.The relevant information contained in the spectral data was fully utilized for effective origin identification,and a method for qualitative identification of soybean origin based on THz frequency-domain spectroscopy was proposed.The experimental results showed that THz frequency-domain spectroscopy after interval partial least squares(i PLS)spectral selection combined with chemometrics identification of soybeans from three typical origins(Argentina,the United States and China)was feasible.After i PLS and auto scaling pre-processing,the artificial bee colony algorithmsupport vector machine(ABC-SVM)soybean origin identification model had a total identification accuracy of 94.74%.This indicated that after i PLS spectral selection and suitable spectral pre-processing,soybean origin could be identified by THz frequencydomain spectroscopy combined with chemometrics.(3)Research on the quantitative detection model and optimization algorithm of soybean protein based on THz absorption spectroscopy.The THz absorption spectra of the experimental samples were analyzed,and the relationship between the protein content in soybean and THz absorption spectra was studied.Compared with traditional protein detection methods,the quantitative detection model of soybean protein based on THz absorption spectra was established.Afterwards,different data dimensionality reduction algorithms combined with THz absorption spectra were used to reduce the detection time of the protein quantitative detection model.At the same time,it was ensured that the accuracy of the verification results of the quantitative detection model remained unchanged or had a certain improvement.The experimental results showed that after appropriate spectral pre-processing and protein quantitative detection modeling algorithms,THz absorption spectroscopy could be used to detect protein content in soybeans.After standard normal variate(SNV)pre-processing combined with the second derivative,the artificial bee colony algorithm-support vector regression(ABC-SVR)protein quantitative detection based on THz absorption spectroscopy showed the best prediction results.Its related coefficient of prediction set(Rp),the root mean square error of prediction set(RMSEP),and the relative standard deviation(RSD)of the prediction set were respectively 0.9659,1.3085%,and 3.5334%.After the suitable data dimensionality reduction algorithm,it was feasible to use THz absorption spectroscopy to quickly and accurately detect the soybean protein.Compared with spectral pre-processing,the data dimensionality reduction algorithm combined with THz absorption spectrum could quantitatively detect protein in soybeans faster and more accurately.The Rp,RMSEP,RSD and calculation time of the back propagation neural network(BPNN)quantitative detection model combined with linear discriminant analysis(LDA)were respectively 0.9677,1.2467%,3.3664%,and 53.51 s.This indicated that this method was of great significance for the rapid and accurate quantitative detection of certain in agricultural products and food by THz absorption spectroscopy.(4)Preliminary research on the quantitative detection of acid value and peroxide value in soybean based on THz absorption spectroscopy.The quantitative detection model of soybean acid value and peroxide value based on THz absorption spectroscopy and data dimensionality reduction algorithms were established.The prediction results of each quantitative detection model for acid value and peroxide value were analyzed,and the possible causes of problems in the verification results were explained.The best quantitative detection model and data dimensionality reduction algorithm for acid value and peroxide value were respectively found.The experimental results showed that after dimensionality reduction by LDA,the BPNN soybean acid value and peroxide value quantitative detection model based on THz absorption spectrum could still achieve the best verification results.The best related coefficient of correction set(Rc),Rp and RMSEP of soybean acid value quantitative detection model were respectively 0.8029,0.7421 and 0.3605 mg/g.The best Rc,Rp and RMSEP of the peroxide value quantitative detection model were respectively 0.8945,0.7633 and 0.5297 mmol/Kg.Although the soybean acid value and peroxide value quantitative detection model based on THz absorption spectrum could quickly detect the acid value and peroxide value in soybean,the accuracy of the quantitative detection model needed to be further enhanced.This method had certain reference value for the quantitative detection of acid value and peroxide value in agricultural products and food by THz absorption spectroscopy.(5)Research on the selection method of characteristic spectrum for the quantitative detection of vitamin E in soybean based on DFT and THz absorption spectroscopy.The feasibility of simulating the absorption peak position of a certain chemical composition based on DFT and THz absorption spectrum was analyzed.At the same time,the simulation level parameter and the actual error of the absorption peak position were determined.Secondly,the absorption peak positions of vitamin E(?,?,?-tocopherol)in soybeans were simulated by DFT and THz absorption spectroscopy.After that,the selection method of characteristic spectrum for the quantitative detection of vitamin E in soybeans based on the absorption peak positions was explored.Finally,the quantitative detection model was established,and the results were verified and analyzed.The experimental results showed that it was feasible to simulate the absorption peak position of a certain chemical composition by DFT and THz absorption spectroscopy.At the same time,it was also determined that the simulation level parameter is B3LYP/6-31+g(d,p)and the actual relative error of the absorption peak position was 2.3155%.The absorption peak positions of vitamin E in soybean were simulated by DFT and THz absorption spectroscopy,and it was found that there were four obvious absorption peaks in the 0.1-1.5 THz frequency band.The absorption peak positions were respectively0.8862,0.9367,1.0296 and 1.1429 THz.The THz absorption spectrum of soybeans were selected by the method(the selection method of characteristic spectrum for the quantitative detection of vitamin E in soybean based on the absorption peak positions)proposed in this paper.Finally,after second derivative pre-processing,the Rp and RMSEP of the ABC-SVR soybean vitamin E quantitative detection model were respectively 0.8241 and 1.3562 mg/g.Although the accuracy of the verification result had been improved to some extent,its accuracy needed to be improved if it was to be applied in practice.Even so,this method was still helpful to the related research on the quantitative detection of a certain chemical composition(with a specific molecular formula)in the mixture.This thesis focused on the identification of transgenic and non-transgenic soybeans,and proposed a method based on THz frequency-domain spectroscopy combined with the spectral region optimization algorithm to quickly and accurately identify transgenic and non-transgenic soybeans;In order to solve the problem of soybean origin identification,a new method of soybean origin identification based on THz frequencydomain spectroscopy was proposed;For the problem of protein quantitative detection in soybeans,a quantitative detection method of protein in soybeans based on THz absorption spectroscopy was proposed.To further reduce the quantitative detection time by combining data dimensionality reduction algorithms,while ensuring that the accuracy of the quantitative detection model validation results remained unchanged or was somewhat improved;To the problem of quantitative detection soybean acid value and peroxide value,the THz absorption spectrum of the experimental sample and data dimensionality reduction algorithms were taken as the starting point,and a preliminary research was carried out on the quantitative detection of soybean acid value and peroxide value based on THz absorption spectroscopy;For the quantitative determination of vitamin E in soybean,the feasibility of simulating the absorption peak position of a chemical component based on DFT and THz absorption spectroscopy was first analyzed using tartaric acid as an example.Meanwhile,the simulation level parameter and the actual error of the absorption peak position were determined.Then,the absorption peak positions of vitamin E in soybeans were simulated by DFT and THz absorption spectra.Finally,the selection method of characteristic spectrum for the quantitative detection of vitamin E in soybean based on the absorption peak positions was proposed,and the quantitative detection model was established,validated and analyze.