摘要
基于烟叶近红外光谱、Heracles电子鼻及二者的融合数据,建立了云南、河南、福建和吉林4个省份的烟叶产地识别模型以及河南省内漯河、南阳、平顶山、许昌和驻马店5个地级市的烟叶产地识别模型.对于地理位置相距比较远的不同省份的烟叶,基于单一数据源就可以建立准确率比较高的产地识别模型.对于河南省内5个地级市的烟叶,其地理位置相距近,气候变化小,烟叶相似性高,仅基于单一信息源的数据,该产地识别模型的准确率偏低.为了提高河南省内5个地级市烟叶产地识别的准确率,将烟叶近红外光谱数据与Heracles电子鼻数据进行融合,由于增加了烟叶数据信息量,这5个产地的识别效果明显提升,其留一法内部交叉验证准确率为98.26%,高于数据融合前单一数据源判别模型的86.96%.研究表明Heracles电子鼻数据可以在不同的数据维度上,对近红外光谱数据进行信息量补充,为烟草品种溯源、质量监测、市场监督等方面提供新思路.
In this study,the tobacco leaf origin identification models were established in four provinces in China(Yunnan,Henan,Fujian,and Jilin)and five prefecture-level cities within Henan Province(Luohe,Nanyang,Pingdingshan,Xuchang,and Zhumadian)by utilizing near-infrared spectroscopy data,Heracles electronic nose data,and a fusion of both datasets.In geographically distant provinces,accurate origin identification models with relatively high precision were successfully constructed by using a single data source.However,in the five closely located cities in Henan Province,where geographical proximity,minimal climate variations,and high tobacco leaf similarity were evident,the accuracy of the origin identification model based on a single information source was comparatively lower.To enhance the accuracy of tobacco origin identification in the five prefecture-level cities in Henan Province,a fusion of near-infrared spectroscopy data and Heracles electronic nose data is performed.The increased information content in the fused dataset significantly improved the identification accuracy in these five origin regions.The Leave-One-Out cross-validation accuracy in these regions was measured at 98.26%,surpassing the 86.96%accuracy of the single-data-source discrimination model before data fusion.This study demonstrates the capability of Heracles electronic nose data to complement near-infrared spectroscopy data across different data dimensions,providing new perspectives for tobacco variety tracing,quality monitoring,and market supervision.
作者
汪阳忠
张鑫
蔡振波
黄雯
费婷
吴达
张旭峰
孟祥周
束茹欣
Wang Yangzhong;Zhang Xin;Cai Zhenbo;Huang Wen;Fei Ting;Wu Da;Zhang Xufeng;Meng Xiangzhou;Shu Ruxin(Technology Center,Shanghai Tobacco Group Co.,Ltd.,Shanghai 200082,China;College of Environmental Science and Engineering,Tongji University,Shanghai 200092,China)
出处
《河南师范大学学报（自然科学版）》
CAS
北大核心
2024年第2期104-110,共7页
Journal of Henan Normal University(Natural Science Edition)
基金
国家自然科学基金(42177378)
国家烟草专卖局卷烟烟气重点实验室开放研究基金课题(2021-7)。
