Development of Intelligest Yield Estimation System for Dragon Fruit Orchard Based on Image Processing

Abstract

Dragon fruit, a tropical fruit renowned for its unique appearance and nutritional value, has gained significant popularity in recent years. The development of an intelligent yield estimation system for dragon fruit orchards can have broader implications for the agricultural industry. It can enable data-driven decision-making, improve supply chain management, and enhance market analysis. The system can also be integrated with existing farm management systems, precision agriculture technologies, and decision support systems to provide a comprehensive solution for farmers. The developed system has utilized advanced sensors and controllers to accurately count the number of fruits, measure their size, and calculate the yield. The detection performance was studied on the basis of accuracy, precision, recall, F1 score, detection accuracy and yield estimation accuracy. The results obtained at different speeds of operation viz.  2 km/h, 3km/h and 4 km/h with different deep learning models viz. SSD, YOLOv2 and YOLOv3. The maximum size detection accuracy using SSD, YOLOv2, and YOLOv3 was 94.16 %, 92.69% and 94.74% respectively, observed at 2 km/h operating speed.  The developed yield estimation system can estimate the yield of dragon fruit with an average 93.92 % yield estimation accuracy at 2 km/h operating speed and 70 cm distance of camera from the tree using SSD model.