Research Method For Rapid Identification And Content Detection Of Chinese Herbals Based On Terahertz Spectroscopy

With the strong support of national policies,Chinese medicine occupies an increasingly important position in the country’s medical system.Chinese herbals is the most important part of Chinese medicine raw materials,and its quality is related to the effect of Chinese medicine treatment.However,there are still frequent quality problems in Chinese herbals.One of the key issues to strengthen quality control is how to conduct rapid and effective detection.At present,traditional Chinese herbals detection methods cannot achieve rapid and efficient detection goals.Terahertz time-domain spectroscopy has been a research hotspot in recent years.It provides a feasible solution for the rapid detection of Chinese herbals because of its advantages of fingerprint spectrum,high signal-to-noise ratio,safety and rapidity.Due to the complexity of Chinese herbals,there are many problems that need to be solved when using terahertz time-domain spectroscopy for detection.In this paper,based on terahertz time-domain spectroscopy technology,combined with spectral analysis and processing,machine learning and other related theories and methods,the rapid identification and content detection methods of Chinese herbals are studied,which has important research significance.The specific research contents are as follows:Firstly,the SPCS-SVM classification model is proposed to solve the problem that the similar terahertz absorption spectrum affects the classification accuracy of the model when classifying Chinese herbals of the same type and different grades.Different grades of ginseng and Panax notoginseng were detected by terahertz time-domain spectroscopy,and the corresponding absorption spectra were calculated.In order to improve the speed of classification,Principal Component Analysis(PCA)is used to reduce the dimensionality of the data.Based on support vector machine(SVM),a classification model is established,and the swarm intelligence algorithm is used to optimize the kernel parameters and penalty factors of the model.Aiming at the problem that the conventional optimization algorithm is easy to fall into the local optimum,the SPCS-SVM classification model is obtained by improving the step size and the strategy of balancing the global search and the local search to optimize the cuckoo algorithm(SPCS).The model was used to classify and test the terahertz spectra of ginseng of different years and different grades of Panax notoginseng.The results show that the SPCS-SVM model has high classification accuracy and stability for similar detection objects,which provides an effective method for the rapid classification and identification of Chinese herbals.Secondly,a recognition method of mean shift algorithm combined with PCA is proposed,which solves the problem of establishing a model to improve the recognition speed and accuracy when recognizing unlabeled and confusing Chinese herbals.Codonopsis,Panax ginseng,American ginseng and Bletilla striata from different origins were collected to obtain terahertz absorption spectra.By analyzing the spectral characteristics,the spectral frequency bands with more noise were cut off,and the PCA method was used to reduce the dimensionality of the spectral data.In order to meet the needs of Chinese herbals identification of unlabeled species,an unsupervised clustering model was established.The established clustering algorithm is used for clustering identification on the data of Bletilla striata from different species and different origins.The mean shift algorithm has the highest identification rate and fast identification speed,and the clustering process is less than 0.1seconds.The research results show that based on terahertz time-domain spectroscopy,the mean-shift algorithm combined with the PCA method can quickly and accurately identify unlabeled and confusing Chinese herbals.Finally,the combination method of Monte Carlo Verification Uninformative Variable Elimination(MCUVE)and Support Vector Regression(SVR)is adopted to solve the problem of non-linearity and difficult balance between information extraction and invalid information elimination in the detection of Chinese herbal content,and to improve the accuracy and stability of Chinese herbals content analysis.The incorporation of sucrose and maltose in red ginseng into wheat flour was detected by terahertz time-domain spectroscopy,and absorption spectra of different concentrations were obtained.Due to the existence of non-linear characteristics and useless information in the absorption spectrum independent of content concentration,the PCA method,Monte Carlo verification non-informative variable elimination method and competitive adaptive weighted sampling method were used to screen or process the spectral variables,and a quantitative regression model was established.By verifying the absorption spectrum of mixtures of different concentrations,the results show that the MCUVE-SVR method has the highest correlation coefficient and the lowest root mean square deviation compared with other methods,which is suitable for the quantitative analysis of Chinese herbals.