Shiva Ahmadi - Azarbaijan Shahid Madani University
Ahmad Mani-Varnosfaderani - Department of Chemistry, Tarbiat Modares University, Tehran, Iran
Buick Habibi - Department of Chemistry, Azarbaijan Shahid Madani University, Tabriz, Iran

Nutritional value and quality features of oils are the most important factors that should be considered in food industry. There is no pure edible oil with appropriate oxidative stability and nutritional properties. Therefore, vegetable oils are blended to improve their applications and to enhance their nutritional quality. Characterization of edible oils is important for quality control and identification of oil adulteration. In this work, we propose a simple, rapid, inexpensive and non-destructive approach for characterization of different types of vegetable oil blends according to the corresponding color histograms. Regression models were applied on four datasets of binary edible oil blends including; Palm Olein-Rapeseed, Palm Olein-Sunflower, Soybean-Sunflower and Soybean-Rapeseed. In all of the aforementioned data sets, despite the high performances of Support Vector Regression (SVR) and Levenberg-Marquardt Artificial Neural Network (LMANN) regression models in terms of coefficient of determination, Bayesian Regularized Artificial Neural Networks (BRANN) provided better results up to ۹۷% for HSI color histograms in both the training and test sets. In order to reduce the numbers of independent variables for modelling, principle component analysis (PCA) algorithm was used. Finally, the results of image analysis were compared with those obtained by processing of FT-IR spectra of mixtures of edible oils. The results revealed that image analysis of mixtures of edible oils yield comparable results to those obtained by processing of FT-IR spectra for characterization of edible oils. Our results suggest that the proposed method is promising for characterization of different binary blends of edible oils.

Multivariate calibration, Edible oil analysis, Image histograms, Artificial Neural Networks, Bayesian regularization