Application of Artificial Neural Networks (ANN) and Image Processing for Prediction of the Geometrical Properties of Roasted Pistachio Nuts and Kernels

Abstract

Roasting is the most common way for pistachio nuts processing, and the purpose of that was to increase the products total acceptability. Purpose of this study was to investigate the effect of temperature (90, 120 and 150°C), time (20, 35 and 50 min), and roasting air velocity (0.5, 1.5 and 2.5 m/s) on geometrical attributes of pistachio nuts and kernels including principle dimensions, shape factor, sphericity, surface area, shell splitting, and true volume. An experimental method and image processing were used in order to measure the geometrical properties. The Artificial Neural Networks (ANN) method was used for predicting the correlation between experimental and image properties. The results showed that the time, temperature, and roasting air velocity didn't have significant effect on principle dimensions, shape factor, sphericity, surface area, shell splitting, and true volume. In all cases, the shape factor of pistachio nuts and kernels were more than 1.25. So, pistachio samples had ellipsoid shape. Pistachio kernels had more similarity to ellipsoid shape in comparison with pistachio nuts. The results revealed that ANN could predict the length, width, height, shape factor, sphericity, shell splitting, surface area, and true volume of roasted pistachio nuts and kernels.