|
|||||
|
|
| 红外光谱数据融合对栽培滇重楼产地鉴别* | |
| 引用本文： | 丁于刚,张庆芝.红外光谱数据融合对栽培滇重楼产地鉴别*[J].云南中医学院学报,2022(1):72-78. |
| 作者姓名： | 丁于刚 张庆芝 |
| 作者单位： | （云南中医药大学中药学院，云南 昆明 650500） |
| 基金项目： | 收稿日期： 2021 - 10 - 03 * 基金项目： 云南省重大科技专项（202002AA100007）；云南省中青年学术和技术带头人后备人才项目（202005AC160032） 第一作者简介： 丁于刚（1995-），男，在读硕士研究生，研究方向：中药资源开发与利用。 △通信作者： 张庆芝，E-mail：ynkzqz@126. com |
| 摘 要： | 目的 多样的环境因素使得不同产地栽培滇重楼的化学成分也丰富多样，不同居群栽培滇重楼的甾体皂苷类成分具有很大的差异，多源数据融合分析能更全面的表征药材化学信息，建立一个高效而准确的产地鉴别模型，为其资源合理开发利用提供依据。方法 以来自云南和四川的8个产地（保山、楚雄、大理、红河、丽江、成都、文山、玉溪）共366份栽培滇重楼根茎为实验材料，采集其傅里叶变换近红外光谱（FT-NIR）和衰减全反射-傅里叶变换中红外光谱（ATR-FTMIR）数据。采用Kennard-Stone算法将不同产地的样品分为2/3的训练集和1/3的预测集，基于4种特征变量提取方法（CARS、VIP、SPA、SO-Covsel）结合2种数据融合策略（低级数据融合和中级数据融合），建立偏最小二乘产地判别分析模型。根据模型参数交叉验证均方根误差（RMSECV）和预测均方根误差（RMSEP）评估模型的稳定性，模型训练集和预测集准确率（ACC）评估模型分类性能。结果 近红外光谱和中红外光谱均能反应不同产地栽培滇重楼的化学成分差异，在中级数据融合中，基于VIP和SPA提取的特征变量建立的模型正确率均大于94%。相较于中级数据融合，低级数据融合模型得到了最为满意的结果，其预测集分类正确率达到100%。结论 根据近红外和中红外数据建立的低级数据融合PLS-DA模型，能够用于栽培滇重楼的产地鉴别分析。 |
| 关 键 词： | 栽培滇重楼 产地鉴别 偏最小二乘判别分析 特征变量提取方法 数据融合 |
The Origin Identification Study of Paris Polyphylla var. Yunnanensis Based on the Infrared Spectrum Data Fusion Strategy |
|
| DING Yugang,ZHANG Qingzhi.The Origin Identification Study of Paris Polyphylla var. Yunnanensis Based on the Infrared Spectrum Data Fusion Strategy[J].Journal of Yunnan College of Traditional Chinese Medicine,2022(1):72-78. | |
| Authors: | DING Yugang ZHANG Qingzhi |
| Affiliation: | ( College of Traditional Chinese Medicine, Yunnan University of Chinese Medicine, Kunming 650500, China) |
| Abstract: | Objective Various environmental factors make the chemical composition of cultivated Paris polyphylla var. yunnanensis also rich and diverse, and the steroidal saponins of Paris polyphylla var. yunnanensis in different origins are quite different. An efficient and accurate identification model of different origins is established to provide a basis for the rational development and utilization of its resources. Methods A total of 366 rhizomes from 8 origins in Yunnan and Sichuan provinces (Baoshan, Chuxiong, Dali, Honghe, Lijiang, Chengdu, Wenshan, Yuxi) were used as experimental materials, and their Fourier transform near-infrared spectra (FT-NIR) and attenuated total reflection-Fourier transform mid-infrared spectroscopy (ATR-FTMIR) data were collected. The Kennard-Stone algorithm was used to divide samples from different origins into 2/3 training set and 1/3 prediction set. Based on 4 feature variable extraction methods (CARS, VIP, SPA, SO-Covsel) combined with 2 data fusion strategies (Low-level data fusion and intermediate-level data fusion), the partial least squares origin discriminant analysis model was erected. According to the model parameters, cross-validation root mean square error (RMSECV) and prediction root mean square error (RMSEP) were used to evaluate the performance of the model, and the accuracy of training set and prediction set were used to evaluate the model classification performance. Results Both near-infrared and mid-infrared spectra could reflect the differences of cultivated Paris polyphylla var. yunnanensis in different origins. In the mid-level data fusion models, the accuracy of the models established based on the feature variables extracted by VIP and SPA were both greater than 94%. Compared with the mid-level data fusion models, the low-level data fusion model obtained the most satisfactory results, and its accuracy reached 100%. Conclusion The low-level data fusion PLS-DA model established based on the near-infrared and mid-infrared data can be used for the identification and analysis of the cultivated Paris polyphylla var. yunnanensis. |
| Keywords: | cultivated Paris polyphylla var yunnanensis geographical authentication PLS-DA analysis feature variable extraction method multi-blocks data fusion strategy |
| 点击此处可从《云南中医学院学报》浏览原始摘要信息 | |
| 点击此处可从《云南中医学院学报》下载全文 | |
