Research On THz Spectral Characteristics And Safety Detection Of Grains

With the in-depth study of spectral resolution and perspective of THz wave,people gradually realize that THz technology has unique application value in the qualitative and quantitative detection of grain quality.The THz spectrum of grains has very rich physical and chemical information,and its unique fingerprint spectrum provides scientific basis for the application of THz technology in grain detection.In this paper,the qualitative identification and quantitative analysis methods of the cereals were studied systematically based on the transmissive THz-TDS system,it provided a new approach and important technical support for the development of grain quality testing.Through the experiment and analysis,the following results are obtained.?1?The basic methods for the extraction of absorption and refractive index of solid and liquid samples,and dielectric constants of dielectric materials by commonly used transmission THz-TDS systems are described.Based on the theory of on-chip antenna orthogonal polarization,a novel method is proposed for determining dielectric constant?and conductivity?of dielectric materials by THz detection device.?2?The THz spectral of the rice,maize,millet,black rice were investigated in the range of0³THz by THz-TDS system.The results show that the characteristic absorption peak and absorption intensity are significant differences,the absorption peak of rice is located 0.20,0.49,0.71THz;maize is located 0.33,0.54,0.74THz;millet is located 0.38,0.60,0.83THz;black rice is located 0.22,0.45,0.76THz,and the refractive index generally meets9)_{?77??6?86?4??80?}>9)₄?80?)>9)₈4)7)7)0))>9)₈?64?0)).The absorption in the frequency range of 0¹.6THz is preprocessed by Savitzky-Golay smoothing,and the PLS-DA model is constructed after PCA processing,the accuracy of the training set and test set was 97.29%and 95.94%,respectively.?3?The terahertz spectral characteristics of normal,worm-eaten and moldy millet grains were measured by THz-TDS system,and combining absorption with pattern recognition algorithm to identify the quality of millet.The study found that the absorption coefficient of different qualities millet are more distinguishable,the absorption peaks of normal,worm-eaten and moldy millet are located at 0.373,0.621THz;0.362,0.603THz;0.355,0.59THz respectively,and the refractive index generally meets9)₈7)?9?<9)₈-0)?60?9))<9)₉8)?67?).We built the PLS-DA model and compare the effects of PLS-DA model using multiple pretreatment and characteristic wavelength extraction methods to obtain the best pretreatment and feature wavelength extraction.The 16characteristic wavelengths are selected on the basis of this best method,and combining an improved particle swarm optimization?IPSO?method based on radial basis kernel?RBF?-based SVM and 10-fold cross validation to build an IPSO-SVM model,the results are compared with other models?PLS-DA,LS-SVM,and PSO-SVM?,it's show that the IPSO-SVM model is superior to other models in predicting performance and computing speed,the accuracy rate of training set and test set was 100%and 97.78%,respectively.?4?The THz spectral of flour,talcum powder and their compounds were analyzed in range of0³THz using THz-TDS system at room temperature,and analyze the difference of THz spectral characteristics.The talcum powder has obvious characteristic absorption peak at 1.14 THz,the absorption coefficient of the mixed sample decreases with the increase of its content,and the absorption peak becomes more obvious and steep.The refractive index increases slowly with the increase of content,and there is a crossover phenomenon in the spectral lines,there is a significant resonance phenomenon above the 1.6 THz.The methods of iPLS,SiPLS,BiPLS,and MwPLS spectral interval selection were applied to quantitative analysis of THz spectral.The study indicated that the models obtained from the four spectral interval selection were superior to the full-spectrum PLS model,and the MwPLS model had the best prediction performance,RMSE_CVV and R_CV of training set were 1.3532 and 0.9972 respectively,theRMSE_P and _PR of test set were 1.9133 and 0.9926,respectively.And the detection limit of talcum powder concentration in mixtures is 1%.The method of spectral interval selection eliminates spectral information effectively that is irrelevant to talcum powder content and optimizes spectral information with greater correlation,THz-TDS technology combined with spectral interval selection could be successfully applied to determine talcum powder in flour,and it has potential application in the fields of detection of illegal additives.The spectral region screening can effectively eliminate spectral information that is irrelevant to the content of talcum powder and optimize correlation.The precision and complexity of the model detection are greatly improved,and a fast and efficient method for the quantitative detection of flour additives can be obtained.