IDENTIFICATION OF DEFECTIVE CHERRIES USING CONVOLUTIONAL NEURAL NETWORK

Worldwide, 2.45 million tons of cherries were produced in 2017, 2.57 million tons in 2018 and 2.59 million tons in 2019. It is very important to sort defective sweet cherries to ensure that the export capacity of sweet cherry producing countries is high. Because defects in fruits are contagious. A single rotten sweet cherry can cause all sweet cherries to rot. Therefore, a model was proposed in the present study to prevent the spread of decay. Defective and non-defective sweet cherry images are classified in the proposed model. In developing countries, defective fruits are sorted manually during or after harvest of fruits. Checking defective fruits during harvest requires a special effort and is time consuming. This, in turn, increases the cost of labor. The cost of labor are reduced with the proposed model. A data set consisting of 1,050 images with a resolution of 224x224 pixels was created in the study. Convolutional Neural Network (CNN) was used for the extraction of characteristics and SOFMAX was used for classification. Other methods were also examined using the transfer learning approach to compare the performance of the proposed system. The highest success rate in the classification of sweet cherries was achieved with the proposed method.