The feasibility of using image processing and artificial neural network for detecting the adulteration of sesame oil

Abstract

Sesame oil is one of the most important nutritional and economic products in the world, which yields a high economic value for many countries every year. The production of sesame oil due to the complex process is very costly and this will increase the price of the product. High prices of sesame oil provide the condition and motivate profiteers to misuse this situation. The main method of misuse is to mix oils such as corn, canola, sunflower, and soybean oil into pure sesame oil. Therefore, the main goal of this study is to evaluate a portable intelligent system for detecting oil adulteration using the technology of machine vision. For evaluation of the adulterated sesame oil, 13 samples of oilseed, sunflower and canola, which included different percentages of their mixing, were prepared. In order to detect the adulteration of sesame oil, artificial neural networks were designed and evaluated with a hidden layer with a number of different neurons (from 1 to 20). The best network with structure 12-6-7 provided for the prediction of mixed samples of sesame oil and sunflower seeds, correlation coefficient (R) and mean square error (MSE) of 0.944 and 0.006, respectively. Also, the best network with structure 12-10-7 to predict the mixed samples of sesame oil and canola provided correlation coefficient and mean square error of 0.946 and 0.0003, respectively. These results show the success of color modeling for detecting of adulteration in pure sesame oil.