This work proposes the classification of oils according to its category and suppliers through data collection and multivariate analysis comparing two analytical methods: digital image capture and UV spectroscopy. Two groups of soybean and corn oils produced by four different suppliers were analyzed with UV spectroscopy with wavelength ranging from 190 to 310 nm, and digital images were obtained using dedicated camera in Raspberry Pi. The color frequency distributions were red (R), green (G), blue (B), grayscale (Gr), hue (H), brightness (Br), and saturation (S). The classification models applied were linear discriminant analysis with a selection of variables by a stepwise algorithm (LDA/SW), partial least squares discriminant analysis (PLS-DA), and support vector machine (SVM). The correct classification rate using the SVM model and GrRGBHBrS data in terms of its suppliers was 98.9%, being the digital image capture equal or better than UV spectroscopy for soybean and corn oils, respectively. The application of digital images enables oil classification from different suppliers and can be considered an efficient method for counterfeiting detection. In comparison with classical methodologies such as chromatography and UV spectroscopy, this new technology is beneficial due to the low cost, fast analysis, being a non-destructive approach, and virtually no sample preparation required.

这项工作通过比较两种分析方法：数字图像捕获和紫外光谱，通过数据收集和多元分析，提出了根据油类和供应商分类的油类。使用紫外光谱法分析了四个不同供应商生产的两组大豆和玉米油，其波长范围为190至310 nm，并使用Raspberry Pi中的专用相机获得了数字图像。颜色频率分布为红色（R），绿色（G），蓝色（B），灰度（Gr），色相（H），亮度（Br）和饱和度（S）。所应用的分类模型是通过逐步算法（LDA / SW）选择变量的线性判别分析，偏最小二乘判别分析（PLS-DA）和支持向量机（SVM）。使用SVM模型和供应商的GrRGBHBrS数据得出的正确分类率为98.9％，分别是大豆油和玉米油的数字图像捕捉等于或优于UV光谱。数字图像的应用实现了来自不同供应商的油品分类，可以被视为一种有效的伪造检测方法。与传统方法（例如色谱法和紫外光谱法）相比，该新技术具有成本低，分析速度快，无损检测的优点，并且几乎不需要样品制备。数字图像的应用实现了来自不同供应商的油品分类，可以被视为一种有效的伪造检测方法。与传统方法（例如色谱法和紫外光谱法）相比，该新技术的优势在于成本低，分析速度快，无损检测，并且几乎不需要样品制备。数字图像的应用实现了来自不同供应商的油品分类，可以被视为一种有效的伪造检测方法。与传统方法（例如色谱法和紫外光谱法）相比，该新技术的优势在于成本低，分析速度快，无损检测，并且几乎不需要样品制备。