Next-Generation Tomato Harvesting: An Intelligent System for Ripeness Determination and Yield Maximization

Engineering notes

Key topics: autonomous driving, object detection. See the paper for implementation details and experimental results.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为端到端自动驾驶、BEV感知、3D目标检测、轨迹预测、路径规划、LiDAR感知等高价值论文补充中文说明。

Original abstract

Harvesting tomatoes is an important aspect of agricultural output, and contemporary farming practices place an increasing demand on precise and effective harvesting techniques. In this research, we present an intelligent system for tomato harvesting, combining machine learning and deep learning approaches for ripeness determination and production optimisation. The YOLOv5 and YOLOv8 object detection models are employed in our approach to identify tomatoes in images and classify them into five distinct maturity stages: tomato_half_ripe, tomato_overripe, tomato_ripe, tomato_rotten, and tomato_unripe. The system is trained and evaluated using a selected dataset provided from Roboflow Universe, containing annotated tomato pictures for model training and validation. Experimental results demonstrate the utility of our technique, with both YOLOv5 and YOLOv8 models getting outstanding accuracy in tomato recognition and classification tasks. Studies comparing the two systems' performance reveal only minor differences, with YOLOv8 doing somewhat better when it comes to identifying tomatoes at more advanced stages of ripeness. Our work sheds light on the potential uses and future developments of deep learning-based precision farming techniques, helping to build intelligent agricultural systems for automated tomato harvesting. Future research may concentrate on integrating multisensor data and further optimising model architectures to increase the effectiveness and dependability of tomato harvesting systems.

Links and sources

• PDF from original source

Comments