AUTHOR=Li Xiaolong , Kong Wenwen , Liu Xiaoli , Zhang Xi , Wang Wei , Chen Rongqin , Sun Yongqi , Liu Fei 

TITLE=Application of Laser-Induced Breakdown Spectroscopy Coupled With Spectral Matrix and Convolutional Neural Network for Identifying Geographical Origins of Gentiana rigescens Franch

JOURNAL=Frontiers in Artificial Intelligence

VOLUME=Volume 4 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/artificial-intelligence/articles/10.3389/frai.2021.735533

DOI=10.3389/frai.2021.735533

ISSN=2624-8212

ABSTRACT=Accurate geographical origin identification is of great significance to ensure the quality of traditional Chinese medicine (TCM). Laser-induced breakdown spectroscopy (LIBS) was applied to achieve the fast geographical origin identification of wild Gentiana rigescens Franch (G. rigescens Franch). However, LIBS spectra with too many variables could increase the training time of models and reduce the discrimination accuracy. In order to solve the problems, we proposed two consecutive variable selections. The first one is to remove variables with near-zero standard deviation to eliminate noise signals. The second one is to retain feature variables iteratively closely related to the geographical origin by variable importance measured (VIM) of random forest. After two selections, the number of LIBS spectral variables was reduced by 98.5% and 98.7% respectively for the underground and aerial parts of G. rigescens Franch. The accuracy of the discriminant models showed an upward trend. The optimal accuracy in the prediction set was obtained after the second variable selection based on convolutional neural network (CNN) with the value of 92.19% and 94.01% for underground and aerial parts of G. rigescens Franch respectively. Moreover, the training time of discriminant models showed a downward trend. Based on underground parts, it took only 312.2 s and 161.3 s for the training of SVM and CNN after the second selection. The present results demonstrate that LIBS combined with variable selection and CNN can be a powerful and efficient tool for the rapid identification of the geographic origin of G. rigescens Franch.