三维荧光光谱法快速检测橄榄油中掺假廉价油

Rapid detection of adulterated cheap oil in olive oil by three-dimensional fluorescence spectrometry

中文关键词:  橄榄油  掺假  三维荧光光谱  机器学习  快速检测

英文关键词:olive oil  adulterate  three-dimensional fluorescence spectrometry  machine learning  rapid detection

中文摘要:

      目的 建立三维荧光光谱法结合机器学习快速检测橄榄油中掺假廉价油的方法。方法 采集橄榄油及掺入大豆油、玉米油、棕榈油3种不同浓度梯度橄榄油的荧光光谱, 采用标准差标准化(Standardscaler)、标准正态变换(standard normal variate, SNV)、归一化(Normalize) 3种光谱预处理方法进行数据处理。基于K近邻(K-nearest neighbor, KNN)、随机森林(random forest, RF)、支持向量机(support vector machine, SVM)、偏最小二乘法(partial least squares, PLS)和卷积神经网络(convolutional neural network, CNN)构建5种橄榄油定量掺假模型。结果 在定性模型中, 基于PLS算法构建的模型效果最好, 对3种掺假橄榄油的判别准确率为0.86~1.00。在定量模型中, Standardscaler预处理结合RF算法构建的模型表现最优, 校正集相关系数、预测集相关系数、校正集均方根误差、预测集均方根误差最高, 分别为1.00、0.99、0.01、0.02。结论 该方法快速、实时、低成本, 适用于橄榄油的定量掺假检测, 为橄榄油质量评估提供方法参考。

英文摘要:

      Objective To establish a method for rapid detection of adulterated cheap oil in olive oil by three-dimensional fluorescence spectrometry combined with machine learning. Methods The fluorescence spectra of olive oil and 3 kinds of olive oil mixed with soybean oil, corn oil and palm oil were collected, 3 kinds of spectral pretreatment methods, Standardscaler, standard normal variate (SNV) and Normalize, were used for data processing. Based on K-nearest neighbor (KNN), random forest (RF), support vector machine (SVM), partial least squares, partial least squares (PLS) and convolutional neural network (CNN) were used to construct 5 kinds of olive oil quantitative adulteration models. Results Among the qualitative models, the model built based on PLS algorithm had the best effect, and the identification accuracies of the 3 kinds of adulterated olive oils were 0.86~1.00. Among the quantitative models, the model constructed by Standardscaler pretreatment combined with RF algorithm had the best performance, and the correction set correlation coefficient, prediction set correlation coefficient, correction set root mean square error and prediction set root mean square error are the highest, which were 1.00, 0.99, 0.01 and 0.02, respectively. Conclusion The method is fast, real time and low cost, which is suitable for the quantitative adulteration detection of olive oil, and provides a method reference for the quality assessment of olive oil.