| 基于色谱和光谱数据融合的不同植物源食用油判别方法与模型 |
| 点此下载全文 |
|
| 基金项目:国家重点研发计划项目( 2017YFE0115400) |
|
| 中文摘要:利用气相色谱和近红外光谱技术对不同植物源的4种食用油(葵花籽油、大豆油、玉米油和花生油)进行表征分析，基于表征数据分别建立了偏最小二乘判别分析(PLS-DA)模型，并在此基础上探究了数据级数据融合方法，构建了基于色谱和光谱数据融合的不同植物源食用油判别方法与模型。主成分分析(PCA)结果显示，气相色谱判别分析主要是依据脂肪酸组成信息，近红外光谱主要是基于样本中含氢化学键的表征进行分类。数据融合模型的灵敏度和特异度均为1000，分类误差为0000，降低了交互验证的平均分类误差，模型具有良好的稳健性。与基于单一数据的模型结果相比，数据融合分析策略提高了模型的分类精度和鲁棒性。 |
| 中文关键词:食用油 植物源 判别分析 气相色谱 近红外光谱 数据融合 |
| |
| Experimental Techniques and MethodsDiscriminant Analysis on Edible Oils of Botanical Origins Based on Data Fusion of Gas Chromatography and Near Infrared Spectroscopy |
|
|
| Abstract:Four kinds of edible oils,ie.sunflower oil,soybean oil,corn oil and peanut oil from different botanical origins were characterized by gas chromatography(GC) and near infrared spectroscopy(NIR),and the discriminant analysis models were established based on the characterization data.Meanwhile,the feasibility of data level data fusion was explored.A partial least squares discriminant analysis(PLS-DA) model was constructed based on chromatographic and spectral data fusion to classify the edible oils of botanical origins.Principal component analysis(PCA) results showed that the discriminant analysis by GC was mainly based on fatty acid composition,while that by NIR was mainly based on the characterization of hydrogen contained chemical bonds in samples.The sensitivity and specificity of the data fusion model were both 1.000,and the classification error were 0.000.Thus,low level data fusion reduced average cross validation(CV) classification error,and the model exhibited a good robustness.Compared with the results of the model based on single data of gas chromatography or near infrared spectroscopy,the data fusion strategy improved the classification performance of the model. |
| Key Words:edible oil botanical origins discriminant analysis gas chromatography(GC) near infrared spectroscopy(NIR) data fusion |
| 引用本文：高冰，吴鹏飞，许晓栋，杨增玲，刘贤.基于色谱和光谱数据融合的不同植物源食用油判别方法与模型[J].分析测试学报,2020,39(11):1398-1403. |
| 摘要点击次数: 1742 |
| 全文下载次数: 735 |
| 查看全文 下载PDF阅读器 |
|
|
