Pistachio nuts are a type of nut that is widely consumed around the world due to their high nutritional value and pleasant taste. Pistachios are usually sold in their shells, either open or closed. However, closed-shell pistachios are not well received by consumers, resulting in a lower commercial value. It is essential to be able to distinguish between open and closed pistachio shells in order to ensure quality control during production processes and processing. This can be done manually or by using mechanical devices. Manual inspection and categorization of pistachio nuts have traditionally been done by workers, but this process is inefficient in terms of time and money. Mechanical separation of open and closed-shell pistachio can damage the kernels of open-shell nuts due to the needle mechanism used in the sorting process. This study aims to classify pistachio nuts using a machine vision-based system and evaluate its applicability in terms of classification accuracy. The system is evaluated on the Antep pistachio species, which can be distinguished from other pistachio varieties, such as Siirt and Urfa pistachios, based on their shape, size, and taste properties. The machine vision system in this study classifies pistachio nuts into closed and open shell classes in a completely automated manner. In this study, 1,000 Antep pistachio nuts images were obtained and examined, including 500 open and 500 closed nuts. The images were pre-processed and prepared for feature extraction. From the images, a total of 14 color features were extracted. Although the single feature was used, promising classification accuracy rates of 95.6%, 94.8%, and 93.6% from the Random Forest, Support Vector Machine (SVM), and Logistic Regression were achieved, respectively. The performances of classifiers were compared to each other. Almost similar performances were detected. These results demonstrate that the Random Forest classifier is the most effective algorithm for classifying open and closed Antep pistachio nuts.
Read full abstract