Quantitative Analysis And Classification Of Solid Pesticide Mixture Based On THz-I DS

The terahertz (THz) wave occupies the frequency region between the infrared andmicrowave (from0.1to10THz,1THz=1012Hz) in electromagnetic spectrum. It has low energy,and will not harm human health and detection objects, thus it can be used to detect theinflammables and explosives. Besides, a great number of substances’ intramolecular vibrationcaused by group torsion and intermolecular interaction arising by Van der waals bond, hydrogenbond and so on can be detected by THz wave, therefore, THz spectroscopy has bright future inmaterial classification. In recent years, a lot of scholars displayed absorption spectra of manymaterials with obvious spectroscopic characteristics. While THz spectroscopy is increasinglyused in quantitative analysis, not just limited in qualitative research, and some study fruits werepresented by some scholars. However, they seldom presented discussion about spectral regionselection for building regression model. This dissertation concentrates on developing proceduresto identify variables (frequency) that contribute the most useful information, aiming at buildingsuperior regression models for quantitative analysis. In addition, considerable effort also has beendirected towards classifying different pesticides. The major contributions and innovations are asfollows:(1) At first, this dissertation stated the study background, significance and purpose. Besides,the conventional methods of material classification and the superiority and feature of THz wavewere also introduced. In addition, the domestic and foreign development statuses of THztechnique were displayed.(2) The THz-TDS system structure, working principle and key parameters were introduced,next, this dissertation described the optical parameters extraction model in detail, which transferthe time domain signal to absorption and refractive index spectrum.(3) Partial least squares (PLS) was employed to analyze the ingredient of ammoniumsulfamate (AMS) and orange mixtures, and obtained good accuracy. Moreover, interval PLS(iPLS), backward iPLS (biPLS) and moving window PLS (mwPLS) were also absorbed to buildregression models towards two sets of pesticide mixtures (thiabendazole and nitrofen). The PLSmodel is built using the full THz spectrum data, and the other three methods conducted variableselection for pesticide THz spectrum to determine the superior interval through fixing four interval widths. Besides, this dissertation discussed and analyzed the influence of interval widthon model prediction accuracy. The bound limit of solid pesticide in mixture is1.11%.(4) For classification of different bactericides’ THz spectra, wavelet packet was used todecompose the terahertz time domain spectra of four bactericides to calculate the energyspectrum, which was regarded as feature vectors to perform classification through clusteringanalysis, and the accuracy rate of classification is nearly100%for20samples.In the last part of the dissertation, some suggestions about further study of THz spectroscopyin quantitative analysis and material classification were put forward.