Ö. O. Dursun - Avionics Department, Civil Aviation High School, Firat University, ۲۳۱۱۹ Elazig, Turkey.
S. Toraman - Air Traffic Control Department, Civil Aviation High School, Firat University, ۲۳۱۱۹ Elazig, Turkey.
Y. Er - Airframes and Powerplants Department, Civil Aviation High School, Firat University, ۲۳۱۱۹ Elazig, Turkey.
E. Oksuztepe - Avionics Department, Civil Aviation High School, Firat University, ۲۳۱۱۹ Elazig, Turkey.

Despite the developing technology, extraction of Sulfured Dried Apricot (Prunus armeniaca) (SDA) stones is still done manually and thus requires a significant amount of labor and time and also causes serious problems in terms of hygiene. According to International Food Standards (CXS ۱۳۰-۱۹۸۱) and Turkish Standard ۴۸۵, the SDA stones must be extracted from the peduncle side of the apricot. Therefore, the correct position of the apricot peduncle and style side must be determined. In this study, a deep learning architecture was improved for the first time to determine the position of SDA stones as a component of the agricultural machine developed to extract SDA stones. In this study, a new Capsule Network architecture was used. With the original capsule network, SDA images were classified with ۸۶.۲۳% accuracy, while it increased to ۹۴.۴۷%with the improved capsule network. Also, the processing time of the developed network architecture was about twice as fast as the original. The result clearly demonstrates that the SDA stone positions are easily determined. Therefore, the designed agricultural machine can extract the SDA stones hygienically and rapidly, without any need for human power.

Capsule networks, Deep features, Prunus armeniaca, Sulfured dried apricots.