Study On The Identification Of Polygonati Rhizoma Based On Hyperspectral Imaging Techbology

According to the Chinese Pharmacopoeia,Polygonatum sibiricum Red,Polygonatum cyrtonema Hua and Polygonatum kingianum Coll.et Hemsl.are derived from dried rhizome,has good clinical application value and food health effect.Polygonatum cyrtonema Hua is mainly distributed in southwest,East and central China,Polygonatum sibiricum Red.is mainly distributed in East China,Northwest,northeast and North China,and Polygonatum kingianum Coll.et Hemsl.is mainly distributed in Yunnan,Guizhou,Sichuan and some surrounding areas.The appearance,internal quality and efficacy of Polygonati Rhizoma from different producing areas are different.It is beneficial to distinguish Polygonati Rhizoma from different producing areas for tracing and quality control of Polygonati Rhizoma.In addition to the 3 kinds of Polygonati Rhizoma prescribed in Pharmacopoeia,P.filipes Merr.is often used as a mixed product of P.filipes Merr.in southern China,and its characters are very similar to those of Polygonatum cyrtonema Hua,which is not easy to distinguish.Besides,there are also quality differences among different species of Polygonatum cyrtonema Hua,Polygonatum sibiricum Red.,Polygonatum kingianum Coll.et Hemsl.and P.filipes Merr.Therefore,it is beneficial to trace the origin of Polygonati Rhizoma and maintain market order to accurately identify Polygonati Rhizoma by hyperspectral imaging technology.Since ancient times,the mainstream processing method of Polygonati Rhizoma aureus has been "nine steaming and nine drying".For each stage of "nine steaming and nine drying",there is no uniform specification and standard for the degree of steaming and drying,and the judgment results are not the same only based on experience.Therefore,an objective and accurate criterion is needed to quickly distinguish The Times of "nine steaming and nine drying" of Polygonati Rhizoma aureus.In this study,the first derivative(FD),second derivative(SD),S-G smoothing(S-G),standard normal transformation(SNV),multiple scattering correction(MSC),FD+S-G,SD+S-G were used to preprocess the spectral data.Continuous projection algorithm(SPA)and competitive adaptive reweighting algorithm(CARS)were used to screen the feature bands.Random forest(RF),partial least squares discrimination(PLS-DA)and linear classification support vector machine(Linear SVC)were used to identify the origin,base,nine evaporation and nine evaporation times of radix anthracis.Specific research contents and results are as follows:1.Under the condition of distinguishing the basen,the origin identification of Polygonatum cyrtonema Hua,Polygonatum sibiricum Red.and Polygonatum kingianum Coll.et Hemsl at different spatial scales was carried out,and the following conclusions were drawn:(1)The province-region,county-region and township origin identification model based on the all-band is established.The optimal modeling combination is VNIR+SWIR+FD+Linear SVC.The accuracy of the training setand test set of provinceregion,county-region and township origin identification model are 99.97% and 99.82%,respectively.100.00% and 99.46%,99.62% and 98.39%;The best modeling combination is CARS+FD+Linear SVC.Under this condition,the accuracy of the training set and test set of the province-level,county-level and towns-level Polygonatum cyrtonema Hua origin recognition model is 98.59% and 97.05%,respectively.97.79% and 94.75%,90.13% and 87.95%.From province to township,the accuracy of the model is decreased,but the overall classification performance is better,no matter based on full band or feature band.(2)For the identification of province-wide origin,the optimal modeling combination based on the all-band is VNIR+SWIR+SD+S-G+Linear SVC,and the accuracy of the model’s training set and test set are both 100.00%.The optimal modeling combination is CARS+SD+Linear SVC.The accuracy of model training set and test set is 94.75% and 94.50%,respectively.The optimal modeling combination of Polygonatum sibiricum Red.county origin identification model is SWIR+SD+Linear SVC,SWIR+SD+SG+Linear SVC,VNIR+SWIR+FD+Linear SVC,VNIR+SWIR+FD+PLS-DA,VNIR+SWIR+SD+Linear SVC,VNIR+SWIR+FD+S-G+Linear SVC,VNIR+SWIR+SD+S-G+Linear SVC,VNIR+SWIR+SD+S-G+PLS-DA,the accuracy of 8 kinds of modeling combination training set and test set is 100.00%.The optimal modeling combination is CARS+SD+Linear SVC.The accuracy of model training set and test set is 92.25% and 92.00%,respectively.(3)Origin identification model of Polygonatum kingianum Coll.et Hemsl.The optimal modeling combination is VNIR+SWIR+FD+Linear SVC,VNIR+SWIR+SD+Linear SVC,VNIR+SWIR+FD+SG+Linear SVC,VNIR+SWIR+SD+S-G+Linear SVC,The accuracy of model training set and test set established by four modeling combinations is 100.00%.The accuracy of training set and test set of origin identification model of Polygonatum kingianum Coll.et Hemsl.based on feature band were 98.41% and 97.73%,respectively.The optimal modeling combination is VNIR+SWIR+SD+S-G+Linear SVC.The accuracy of training set and testing set of the model is 99.89% and 99.55%,respectively.The optimal modeling combination is CARS+FD+Linear SVC,and the accuracy of model training set and test set is 85.22% and 81.59%,respectively.Without distinguishing the basic origin,the identification of the origin and the authentic famous and excellent producing area of Polygonati Rhizoma in 12 provinces was carried out,and the results were as follows:(1)The identification model of the origin of Polygonati Rhizoma was established based on the all-band.The results of the model established by VNIR+SWIR+SD+Linear SVC were optimal,and the accuracy of the training set and the test set were 99.95% and 99.77%,respectively.The optimal combination of CARS+FD+Linear SVC was used to establish the origin identification model of Polygonati Rhizoma based on the feature band.The accuracy of model training set and test set were 91.98% and 89.66%,respectively.(2)Based on the all-band identification model,the modeling result of VNIR+SWIR+SD+Linear SVC model is the best,and the accuracy of training set and test set are 100.00% and 99.43%,respectively.Based on the characteristic bands,the identification model of the truly famous and excellent producing areas of Polygonati Rhizoma was established.The optimal combination was CARS+SD+Linear SVC,and the accuracy of the model training set and test set were 91.31% and 90.28%,respectively.2.Identification of Polygonatum cyrtonema Hua,Polygonatum sibiricum Red.Polygonatum kingianum Coll.et Hemsl.and P.filipes Merr.Through the comparison of the identification results of different models,we can see that the identification accuracy of Polygonati Rhizoma is close to 100.00% based on the VNIR+SWIR fusion hyperspectral data after FD pretreatment,and the discrimination model established by Linear SVC can accurately identify Polygonati Rhizoma of different radix.After FD pretreatment,the modeling accuracy of the characteristic bands screened by SPA could reach more than 92.00%.The characteristic bands were the same as the complete bands,and the classification information contained in the characteristic bands could distinguish the flavonoid of different primitive well.3.Based on the hyperspectral data of raw products and nine times of evaporation and drying of Polygonati Rhizoma coptidis,identification of nine times of evaporation and drying of Polygonati Rhizoma was conducted.The results show that after SD+S-G preprocessing of VNIR+SWIR fusion hyperspectral data,the accuracy of the identification model of the nine steaming and nine drying times of the essence of Polygonati Rhizoma is the highest using Linear SVC,which can basically completely distinguish the essence of Huang Jing with different steaming and drying times.The characteristic bands screened by SPA only account for 6.31% of the whole band.After SNV pretreatment,the identification accuracy of the discrimination model of different evaporation times of the Linear SVC algorithm can also reach more than 84.00%.Hyperspectral imaging technology can accurately distinguish the Polygonati Rhizoma of different steaming times,avoiding the subjectivity of traditional experience discrimination.The contents of main chemical components were predicted for the samples of nine steaming and nine drying flavonoids,and the prediction model of total polysaccharide content,total saponins content,total flavonoids content,total phenol content and 5-hydroxymethylfurfural content were established respectively.The results showed that except for the total saponins content prediction model,the other component content prediction models had good prediction effect,and the characteristic bands selected by CARS could replace the whole band to construct the content prediction model,greatly reducing the time of data analysis.