Geographical discrimination and adulteration analysis for edible oils using two-dimensional correlation spectroscopy and convolutional neural networks (CNNs).

Geographical discrimination and adulteration analysis play significant roles in edible oil analysis. A novel method for discrimination and adulteration analysis of edible oils were proposed in this study. The two-dimensional correlation spectra of edible oils were obtained by solvents perturbation and the convolutional neural networks (CNNs) were constructed to analyze the synchronous and asynchronous correlation spectra of the edible oils. The differences for geographical origins of oils or oil types could be amplificated through the networks. For different networks, the layer sequences and the filter number of convolutional layers may affect the analysis results. A group of sesame oils from different geographical origins and a group of olive oils adulterated by other vegetable oils were adopted to evaluate the proposed method. The results show that the proposed method may provide an alternative method for edible oil discrimination and adulteration analysis in practical applications. For the two datasets, the prediction accuracy could be 97.3% and 88.5%, respectively.