Identification of Full and Empty Pistachio Shells Using a Sound Processing Approach in Time and Frequency Domains

Pistachio nuts are the major non-oil export agricultural commodity in Iran. The existence of empty shells among the shells with kernels reduces the quality of the nuts and, consequently, their market value. In pistachio processing plants, empty shells are generally separated from shells with kernels by the flotation method. But this method is not precise and may encourage the growth of fungi in the nuts. This research investigated an identification method based on sound reflection analysis. A sample of pistachio nuts was divided according to size into large, medium and small groups. Each group was further divided into empty and full shells. A sound box and PC computer were prepared to detect and analyze the echoes of the impacts of each of the nuts as it drops onto a steel plate. The impact echoes were recorded and the echo signals analyzed in both time and frequency domains using a nearest distance classification approach. The best results were obtained for the large sized nuts with 98.75% accuracy for detecting empty shells and 82.8% for detecting full shells. The classification results were also good for other classes, with the lowest accuracy rating being 62.75% for medium-sized full shells.