Fundamental And Applied Research Of Detection On Saccharides And Honey Based On Terahertz Time-domain Spectroscopy

As a novel technique,the research both of primary theory and instrumentation on the terahertz technology is still in an early stage.The majority of scholars engaged in the development of terahertz system rather than the application of terahertz waves.There were very few studies on food area and the related references were rare.According to the principle of terahertz waves.terahertz time-domain spectroscopy has show n a good potential in the field of food detection.Therefore,this paper explored the fundamental and applied research of detection of terahertz time-domain spectroscopy on sugar and honey,taken saccharides and honey as the subjects.Then,the basic experiments,perceiving,and analyses were carried out on each step of the detection procedures.The fundamental theory,information collection,optical parameters selection,feature extraction,and model construction were deeply discussed in respectively.This exploration lays a theoretical foundation for the practical detection of terahertz time-domain spectroscopy.The details of research contents and results were as follows:1.Terahertz time-domain spectroscopy is limited by a variety of factors during the information collecting.The elements such as acquisition mode,tablet sample preparation,reference signal,and the numbers of scans were considered.The influence of different matrix materials on the terahertz absorption coefficient spectra of saccharides tablet samples and on the identification models to classify the saccharides were discussed mainh.It was found that PTFE absorbs less terahertz waves and could be used as an alternative to PE,of which the SIMCA model for saccharides identification had better results after pretreatment.Meanwhile.PVC.flour and rice flour could be considered as the candidates for substitute products due to the acceptable results of SIMCA model to differentiate the kinds of saccharides using multiple spectra baseline correction(MSBC).2.Different optical parameters reflected different information of sample.Taking saccharides aqueous solution as the research object different optical parameters such as absorption coefficient.complex refractive index and complex dielectric constant with different concentration and different types of saccharides aqueous solution w ere obtained and discussed.It was found that the absorption coefficient spectra of different saccharides solutions did not exhibit characteristic absorption peaks.The absorption coefficient gradually decreased with increasing concentration.The variation range of the absorption coefficient and the complex refractive index of disaccharide solution with the increase of the concentration was greater than the variation range of the monosaccharide.At 1.0 THz,the linearity between the extinction coefficient and the saccharides solution concentration showed a fitting result.The imaginary part of the complex dielectric constant decreased at 0.3-2.2 THz.and rose at 2.2-3.0 THz,which is consistent with the double Debye model of aqueous solution.3.The three characteristics of absorption coefficient.the slope and area of absorption coefficient spectrum area w ere extracted to establish the single linear regression model(SLR)and multiple linear regression(MLR)model,and then were compared the application of different characteristics in honey authenticity evaluation.It was found that the SLR model and the MLR model considered the area of absorption coefficient spectrum as the characteristics were better than those using the other two features.showing the feasibility of the area of absorption coefficient spectrum used as a feature for the terahertz absorption coefficient spectrum analysis.4.The authenticity of honey was studied based on three data fusion methods consisting of data-level,feature-level and decision-level fusion.It was found that the results of PLS model using data-level fusion parameters was better than that of different single optical parameters.The characteristic frequencies of different single optical parameters were obtained by the stability-competitive adaptive reweighted sampling(SCARS)method and then were used to constructed the model which had the better results than the data-level fusion model and single optical parameter model The prediction results of different single optical parameter model were endowed with weights to fuse in decision-level by means of technique for order preference by similarity to an ideal solution(TOSIS).The decision-level model was worse than the data-level and feature-level fusion model5.Based on the pattern recognition method,the identification of Acacia honey,Vitex honey and Medlar honey was conducted.Terahertz time-domain spectra and absorption coefficient spectra of different flower resources of honey samples were collected.The pattern recognition models,principal component analysis(PCA),cluster analysis(CA),partial least squares discriminant analysis(PLS-DA)and least-squares support vector machine discriminant analysis(LS-SVM-DA),were analyzed.It was found that LS-SVM-DA method was more suitable for terahertz technology to identify honey botanical types.At the same time,successive projections algorithm(SPA)was used to discuss effective information in different frequency bands.It was found that there were more effective information at 0.5-1.5 THz,showing that using the narrow-band terahertz time-domain spectroscopy system and suitable data processing methods could meet the actual application requirements,that under the certain conditions.