The integration of image processing and artificial neural network to estimate four fatty acid contents of sesame oil

Sesame produces prominent oil whose high-quality emanates from its satisfying combination of fatty acids and antioxidants. Oleic, linoleic, palmitic and stearic acids are the four main fatty acids of sesame oil that are presently measured in laboratories using time-consuming and expensive methods that mainly destruct genetic seeds. The aim of this study is to estimate the qualitative traits of oil in sesame seeds from their digital image using image processing techniques. To this end, some seed features including color, length/width ratio, and texture were derived and homogenized from the sesame images. A multilayer perceptron (MLP) artificial neural network was employed to estimate fatty acid contents of 125 different sesame genotypes based on the seed image data. Then, the efficiency of the MLP model was compared against several other modeling techniques. The results showed that the MLP model provided an optimal estimation of all four fatty acids as compared to the other models. In testing, the MLP model phase could estimate linoleic, oleic, stearic and palmitic acids content in sesame oil accurately with a root-mean-square-error of 0.972, 0.619, 0.714 and 0.313 percent and a determination coefficient of 0.938, 0.982, 0.907 and 0.932, respectively.