Partly Interpretable Machine Learning Method of Ginseng Geographical Origins Recognition and Analysis by Hyperspectral Measurements
LI Meng1, 2, ZHANG Xiao-bo2, LIU Shao-bo3, CHEN Xing-feng4*, HUANG Lu-qi5*, SHI Ting-ting2, YANG Rui6, LIU Shu7, ZHENG Feng-jie8
1. School of Pharmacy, Henan University of Chinese Medicine，Zhengzhou 450046，China
2. State Key Laboratory Breeding Base of Dao-di Herbs, National Resource Center for Chinese Materia Medica, Chinese Academy of Chinese Medical Sciences, Beijing 100700, China
3. Big Date Center, Space Star Technology Co., Ltd., Beijing 100086, China
4. State Environmental Protection Key Laboratory of Satellite Remote Sensing, Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, China
5. State Key Laboratory Breeding Base of Dao-di Herbs, Chinese Academy of Chinese Medical Sciences, Beijing 100700, China
6. Key Laboratory of Remote Sensing of Gansu Province, Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, Lanzhou 730000, China
7. Jilin Provincial Key Laboratory of Chinese Medicine Chemistry, Changchun Institute of Applied Chemistry, Chinese Academy of Sciences, Changchun 130022, China
8. School of Space Information, Space Engineering University, Beijing 101416, China
Abstract：Ginseng is a valuable variety of traditional Chinese medicine with high economic value. The growth is very regional, and the effective ingredients of ginseng from different origins are different. Whether ginseng is “authentic” or not, it will cause differences in its quality, medical utility and economic value, so the identification of ginseng origin is of great significance. After powder extraction and other preparations, chemical or optical methods are used to test the origin of ginseng, but this will cause damage to the sample. Besides, the identification based on appearance traits or rhizome head characteristics can not be used as a standardized recognition method because of human subjective differences or easy to be falsified. The main standpoint of this article is how to use high-precision, non-destructive, and rapid detection and identification methods to identify and analyze the origin of ginseng. This experiment uses hyperspectral imaging technology, for ginseng samples with known origin information, the hyperspectral reflectance dataset was constructed by obtaining reflectance spectra from 400 to 2 500 nm, after absolute and relative radiometric corrections based on the whiteboard. A full spectrum ginseng origin recognition model based on hyperspectral data was constructed, and the accuracy of origin recognition was verified for different scales of regional division rules. It was found that the ginseng spectra from different origins were significantly different. The accuracy of origin identification of the northeastern provinces or not can reach 98.2%. The spectral importance results of ginseng origin recognition were given, indicating the characteristic spectrum for developing a special lightweight instrument. As a strict non-destructive detection method, hyperspectral ginseng origin identification research will provide theoretical support and technical means for identifying the origin of authentic Chinese medicinal materials such as ginseng, fingerprint recognition and mining of medicinal materials, identification and quality evaluation, etc.
Key words：Hyperspectral; Random forest; Interpretability; Ginseng; Traditional Chinese medicinal materials; Origin
[1] National Pharmacopoeia Commission（国家药典委员会）. Pharmacopoeia of the People’s Republic of China 2020（中华人民共和国药典2020版）. Beijing：China Medical Science Press（北京：中国医药科技出版社），2020. 8.
[2] BAI Yun-hui, ZHANG Tai-ming, ZHANG Feng-qing（白云慧, 张泰铭, 张凤清）. Science and Technology of Food Industry（食品工业科技），2015，（13）：297.
[3] Cheng C，Yuan Q，Zhou H，et al. Microscopy Research and Technique，2016，79（2）：98.
[4] Chen Y，Zhao Z，Chen H，et al. Journal of Ginseng Research，2017，41（1）：10.
[5] Sung Won Kwon，Sang Beom Han，Il Ho Park，et al. Journal of Chromatography A，2001，921（2）：335.
[6] LUO Zhi-yong，ZHOU Gang，ZHOU Si-qing，et al（罗志勇，周 钢，周肆清，等）. Acta Pharmaceutica Sinica（药学学报），2000，35（8）：626.
[7] Jun Wen, Elizabeth A Zimmer. Molecular Phylogenetics and Evolution，1996，6（2）：167.
