Akademik Veri Yönetim Sistemi
International Conference on Reliable Systems Engineering (ICoRSE) - 2023, Bucharest, Romanya, 7 - 08 Eylül 2023, cilt.762 LNNS, ss.616-621
Automated systems are needed to count the yield in industrial orchards. Computer vision and artificial intelligence techniques offer very promising results for such counting problems. In particular, the use of deep learning methods in outdoor experimental studies, where errors due to the reflection of light are experienced, makes computer vision applications more reliable. In this study, the hyperparameters of the deep learning model applied to autonomously determine the yield in the orchard were optimized and the effect on the result was examined. Apples are determined and counted using the deep learning method on the images obtained from the Multirotor Micro Aircraft (MAVs) developed for this study. To increase the performance of this process, some improvements can be made during the deep learning training phase. Optimizing the hyperparameters is one of these improvements, and it has been observed that it can increase the success performance by up to 30% with the random search method.