The Geographical Discriminatiom Of Amomum Tsao-ko Based On Spectroscopy Combined With Chemometrics

Objective: A.tsao-ko,as a traditional Chinese medicine,is homologous to food and medicine.It has a long history of cultivation and application in China.With the in-depth use of A.tsao-ko,its economic value has gradually increased,which can be considered as an important force in poverty alleviation in Yunnan mountainous areas.A.tsao-ko is widely distributed,and the effective components of A.tsao-ko from different geographical origins are quite different,resulting in uneven quality.In recent years,with the increasing demand for A.tsao-ko in domestic and foreign consumer markets,the demand for discrimination of the geographical origin of A.tsao-ko is gradual growing.Therefore,it is urgent to explore and establish a high-identification,high-speed and effective technical method to distinguish the geographical origins of A.tsao-ko,which is not only the premise of ensuring the quality and quality of A.tsao-ko,but also the focus of the development and research of medicine and food homologous plants.Methods: The A.tsao-ko from different geographical origins were analyzed by near infrared(NIR)and ultraviolet-visible(UV-Vis)spectroscopy.Combined with principal component analysis(PCA),projection variable importance(VIP),partial least squares discriminant analysis(PLS-DA),sequential orthogonal partial least squares(SO-PLS),sequential orthogonal covariance selection(SO-Covsel),sequential preprocessing through orthogonalization(SPORT),red-green-blue image analysis(RGB)and residual convolution neural network(Resnet),the spectral characteristics of A.tsao-ko of different geographical origins were deeply analyzed,and the effects of different chemometrics approaches(pre-processing,variable extraction,data fusion)on model classification were compared to explore a simple,accurate and low-cost method for identifying the geographical origins of A.tsao-ko and to provide a theoretical basis for scientific and standardized utilization of A.tsao-ko resources.Results:(1)the NIR and UV-Vis spectra of 180 A.tsao-ko samples from 5 geographical origins in Yunnan Province were collected.The feature information of the NIR and UVVis were extracted and further fused into a new dataset through two variable reduction strategies(PCA and SO-PLS)and two variable selection strategies(VIP and SO-CovSel).PLS-DA was thus applied to establish the classification model for discriminating the geographical origins of A.tsao-ko.The results showed that SO-PLS and SO-CovSel,taking the advantages of sequential modeling and orthogonalization,could not only identify the common information in the two datasets,but also provide a more concise classification model to identify the origin of A.tsao-ko without losing the classification ability.(2)the NIR spectra of 248 samples of A.tsao-ko from 5 geographical origins in Yunnan Province were collected.First of all,different scattering correction techniques were used to reduce the light scattering in the NIR spectra.The processed spectra combined with the PLS-DA,was thus to establish discriminant models.SPORT,then,was applied to fuse the complementary information from different scattering correction techniques for establishing the discriminant model.The results showed that the SPORT model could use the complementary information related to different scattering correction techniques,which is better than the single scattering correction model.In addition,compared with the traditional data fusion strategy,SPORT can economically and efficiently identify the geographical origins of A.tsao-ko without additional instruments and sample information.(3)the NIR spectra of 439 A.tsao-ko from 6 geographical origins in Yunnan Province were collected and converted into synchronous and asynchronous 2DCOS images.RGB image analysis method was applied to transform 2DCOS image into one-dimensional data matrix.Then,PLS-DA and SVM models were established to identify the geographical origin of A.tsao-ko.Compared with asynchronous 2DCOS images,synchronous 2DCOS images contained more useful information for distinguishing the geographical origins of A.tsao-ko.From the perspective of prediction,the PLS-DA and SVM models based on synchronous 2DCOS images provided the best classification results,which were more suitable for the identification of A.tsao-ko origin.(4)Resnet models,based on synchronous and asynchronous 2DCOS images,were established to identify the geographical origin of A.tsao-ko.Compared with asynchronous 2DCOS images,the Resnet model established by synchronous 2DCOS images could effectively identify the geographical origin of A.tsao-ko,with an accuracy of 100%.In addition,compared with the traditional chemometric methods(PLS-DA and SVM,etc.),Resnet did not need complex data processing and artificial feature extraction,which provided a simple and efficient method for the identification of geographical origins of A.tsao-ko.We hope that the combination of synchronous 2DCOS images and Resnet analysis can provide a useful reference for geographical traceability of the other spices and traditional Chinese medicine.