基于统计学习理论的食用油红外光谱分析方法研究

A study of edible oil infrared spectroscopy method based on statistical learning theory

中文关键词:  食用油  中红外光谱  支持向量机  主成分分析  马氏距离

英文关键词:edible oil  mid-infrared spectroscopy  support vector machine  principal component analysis  Mahalanobis distance

中文摘要:

      目的 建立不同品质食用油快速分类的中红外光谱检测方法。方法 不同品质的食用油在化学组分上是存在差异的, 利用中红外光谱技术全面反映和整体把握食用油的化学成分信息, 并借助主成分分析(PCA)结合马氏距离法对食用油的中红外光谱图进行预处理, 提取其特征信息, 然后通过基于统计学习理论的支持向量机(SVM)建立相应分类模型, 运用模型自动鉴别不同品质的食用油类别属性。结果 实验通过从市场上随机抽取食用油样本, 选取了3种不同品牌的大豆油、花生油共60个样本进行测试, 分类正确率达到了100%。结论 基于统计学习理论的食用油红外光谱分析方法对不同品质食用油的快速分类鉴别是有效的。

英文摘要:

      Objective To establish a rapid classification method of different quality edible oil using mid-infrared spectroscopy. Methods The chemical composition of different quality of edible oil was different, mid-infrared spectroscopy was used to fully grasp and reflect the chemical composition of edible oil information, and the principal component analysis (PCA) combined with Mahalanobis distance method were applied to preprocessing the infrared spectra of edible oil, and extracted feature information, and then appropriate classification model was established by support vector machines (SVM) based on statistical learning theory, the model automatically identify different categories of quality edible oil properties. Results Random edible oil samples were selected from the market, 3 different brands of soya bean oil, peanut oil, and a total of 60 samples were tested, and the correct classification rate was 100%. Conclusion The method based on mid-infrared spectroscopy statistical learning theory is effective for rapid classification and identification of different quality edible oil.