| In recent years,with the continuous expansion of the market of Chinese herbs,the demand for quality inspection of traditional Chinese medicine is growing.Because of the complexity of the component system of Chinese herbs,the traditional analytical method is time-consuming and tedious,and it is difficult to achieve non-destructive and rapid detection of its quality.As an non-destructive and effective analytical method,near infrared spectroscopy(NIRS)has shown great potential in the detection of herbs.At present,most of the near-infrared spectrometers used in the analysis of traditional Chinese medicine ingredients are imported instruments,mainly based on larger Fourier transform near-infrared spectrometers,which are expensive and not suitable for rapid on-site detection.On the other hand,due to the severe overlap of the peaks of the near-infrared spectrum and the wide band,the key to the quantitative analysis of near infrared spectroscopy is to research appropriate data analysis methods to obtain a stable and accurate model.In order to solve the above problems,a portable near infrared spectrometer was designed.The main research object was Angelica dahurica.The analysis of components and identification methods of origin of Angelica dahurica based on near infrared spectroscopy were studied.The main research work of this paper includes the following aspects:Design and testing of portable near infrared spectrometer.In this paper,the optical components and optical structure parameters of the near infrared spectrometer are selected and designed.The simulation and optimization of the optical path are carried out by using Zemax software.The mechanical structure of the components and the whole machine is designed.The software of near infrared spectrum acquisition and data processing is designed based on C# language,which realizes functions such as spectrum acquisition,data storage and loading,spectral overlay display,preprocessing,and model establishment.A prototype of near infrared spectrometer was built and calibrated.The spectrum range is 900～1700nm,the spectral resolution is4 nm,and the external dimensions are 147mm×100mm×60mm,to meet the design expectations.Study on near infrared quantitative method for the content of active ingredient imperatorin in Angelica dahurica.A method for quantitative determination of imperatorin in Angelica dahurica by least squares support vector machine(LS-SVM)is proposed.This method can extract effective information from the spectrum of Chinese herbs effectively and is suitable for the analysis of complex components.Using the content of imperatorin determined by high performance liquid chromatography as the standard value,a quantitative model was established with the sample of calibration set.It was proved that the determination coefficient of the model was 0.988,the calibration root mean square error was 0.009 and the prediction root mean square error was 0.011.The results showed that near infrared spectroscopy could be used for rapid determination of imperatorin in Angelica dahurica with high accuracy.Study on near infrared quantitative method for talc content in Angelica dahurica.A fast and accurate near infrared detection method for talc content in Angelica dahurica based on feature wavelength extraction and LS-SVM is proposed.Moving window method and genetic algorithm are used to select feature wavelength from the original spectrum.Moving window method can locate feature intervals quickly.Genetic algorithm can optimize feature wavelength in the interval.The combination of the two methods is good.Effective and fast screening of wavelengths can improve the prediction accuracy.The quantitative model of LS-SVM is constructed based on the selected wavelength.The anti-interference ability and practicability of the model are verified by the multi-component samples mixed with talcum powder and flour.The determination coefficient of the model is 0.995,the correction root mean square error is 0.42,the predicted root mean square error is 0.48.Compared with the full spectrum model and the least square model,the model established by this method has higher accuracy and smaller error.Study on near infrared identification method for the origin of Angelica dahurica.A near infrared spectroscopy method for the identification of Angelica dahurica origin using random forest pruning algorithm is proposed.This method reduces the complexity of the model and improves the accuracy of the model identification.Near infrared spectra of Angelica dahurica samples from four different regions were pretreated by S-G smoothing,first derivative and SNV.Qualitative discriminant models were established by KNN,SVM and pruning random forest algorithm,respectively.The results show that the identification accuracy of pruned random forest model is higher,and the generalization ability is strong.The accuracy of calibration set is 100% and that of validation set is 98.2 % which provides an effective method for origin identification. |
PDF Full Text Request