The actual storage period of edible oil is one of the important indicators of edible oil quality. A high-precision identification method based on the near-infrared (NIR) spectroscopy technique for the actual storage period of edible oil is proposed in this study. Firstly, a Fourier transform NIR (FT-NIR) spectrometer was used to collect NIR spectra of edible oil samples in different storage periods, and the obtained spectra were pretreated by standard normal transformation (SNV). Then, the characteristics of the pretreated spectra were analyzed by principal component analysis (PCA), and the spatial distribution of edible oil samples in different storage periods was visually presented using a PCA score plot. Finally, three pattern recognition methods, which were K-nearest neighbor (KNN), random forest (RF), and support vector machine (SVM), were compared to establish a qualitative identification model of edible oil in different storage periods. The results showed that the recognition performance of the SVM model was significantly superior to that of the KNN and RF models, especially in terms of generalization performance, and the SVM model had a recognition rate of 100% when predicting independent samples in the prediction set. It is suggested that FT-NIR spectroscopy combined with appropriate chemometric methods is feasible to realize fast and high-precision identification of actual storage periods of edible oil and provided an effective analysis tool for edible oil storage quality detection.

中文翻译：

---

FT-NIR光谱结合化学计量学方法可高精度识别食用油的实际存储时间。

食用油的实际贮存期是食用油质量的重要指标之一。提出了一种基于近红外光谱技术的食用油实际存储期高精度识别方法。首先，使用傅立叶变换近红外光谱仪（FT-NIR）收集了不同储存期食用油样品的近红外光谱，并对所得光谱进行了标准正态变换（SNV）预处理。然后，通过主成分分析（PCA）分析预处理光谱的特征，并使用PCA评分图直观地显示食用油样品在不同储存期间的空间分布。最后，三种模式识别方法，即K比较了近邻（KNN），随机森林（RF）和支持向量机（SVM），建立了不同存储期食用油的定性识别模型。结果表明，SVM模型的识别性能明显优于KNN和RF模型，尤其是在泛化性能方面，并且在预测集中预测独立样本时，SVM模型的识别率为100％。FT-NIR光谱法结合适当的化学计量学方法可以实现对食用油实际存储期的快速，高精度识别，为检测食用油的存储质量提供有效的分析工具。