基于红外光谱的食用植物油种类鉴别Identification of edible vegetable oils based on infrared spectrum

为建立基于红外光谱的食用植物油种类鉴别方法，收集了常见的5种食用植物油样本296份，采集红外光谱，分别通过Savitzky-Golay平滑、希尔伯特变换、IIR低通滤波器、IIR高通滤波器、连续小波变换、一阶导数、二阶导数进行预处理，并利用径向基函数(RBF)神经网络和随机森林(RF)模型对光谱进行识别。结果表明：RBF神经网络模型的效果优于RF模型，将红外光谱数据经希尔伯特变换处理后，RBF神经网络模型的识别率达到100%。采用该方法对食用植物油进行种类鉴别快速无损、准确率高、效果好。 In order to establish the identification method of edible vegetable oils based on infrared spectrum, 296 samples of 5 kinds of common edible vegetable oils were collected, their infrared spectrum were collected and pretreated by Savitzky-Golay smoothing, Hilbert transform, IIR low-pass filter, IIR high pass filter, continuous wavelet transform, first derivative and second derivative respectively, and spectrums were identified by Radial Basis Function(RBF) neural network and Random Forest(RF) models. The results showed that the effect of RBF neural network was better than the RF model. After pretreating the infrared spectrum data by the Hilbert transform, the recognition rate of the RBF neural network model reached 100%. This method has the advantages of rapid non-destructive, high accuracy and good effect in the identification of edible vegetable oils.