Monitoring the maturation of fruits and diagnosis rotten with UAV

Agriculture is considered a fundamental sector of production for humanity, having continuously advanced through technological innovations. Indeed, many countries with economies based on agriculture exist. To enhance agricultural economic progress, modern and technological methods must be employed. As a result, significant increases in agricultural productivity can be achieved, thereby providing substantial support to national economies. Improving the efficiency of fruit production can enhance the quality and quantity of ripe products, influencing consumers' physical health and purchasing motivations. Additionally, it plays a crucial role in determining market prices. This thesis study investigates the effectiveness of the YOLOv8 convolutional artificial neural network model on fruit detection, analysis and counting in the agricultural sector. A total of 2223 images were used as the training set. The study was conducted on classes such as Apple, Orange, Banana and Pomegranate, which include four different types of fruits. The outstanding success of the model is especially evident in its ability to successfully determine fruit ripening stages. The training process took place over 60 cycles, and the learning speed was determined as 2,5 ms. During training, the loss value of the model continuously decreased and the accuracy value increased. As a result of detailed evaluations, the success rates achieved by the model on the basis of fruit types are as follows: 95,% for Apple, 98,% for Orange, 93,8% for Banana and 98.8% for Pomegranate. The overall success rate of the model was determined to be 97,09%, which highlights the model's ability to accurately classify complex and variable fruit characteristics. The findings indicate that the YOLOv8 model can demonstrate reliable performance in fruit detection and classification tasks within the agricultural sector, successfully distinguishing between various fruit types. Furthermore, the experimental analysis of images obtained by Unmanned Aerial Vehicles (UAVs) has highlighted the effective usability of the YOLOv8 model in agricultural applications. The model has achieved success in various agricultural applications, including detection, qualitative analysis, and counting of fruits in different locations. These findings underscore the model's capability to reliably perform fruit detection from UAV images in the agricultural sector, positioning it as an effective tool for agricultural applications.