Tomatoes are one of the most marketed vegetables in the whole world. However, the production of this vegetable has not been maximized due to post-harvest losses. A vital cause of this loss is the speed and accuracy of classifying the maturity level of tomatoes due to the existing slow method available. This study presents an automated system for classifying tomato ripeness and sorting tomatoes using a robotic manipulator and computer vision. The objectives are to build a robotic arm to grip and manipulate tomatoes, develop a computer vision model to classify maturity level, and integrate these to automate the sorting process. The methodology involves assembling a robotic arm with a gripper using servo motors and metal parts. Images of tomatoes were collected to train YOLOv5 and YOLOv8 computer vision models to classify maturity level. The research evaluated the functionality of the robotic arm by assessing movements such as pitch, yaw, and gripping. The vision models were trained on a dataset of tomato images and evaluated for classification accuracy. Comparisons showed YOLOv8 achieved 99.1% precision and 99.5% in detecting ripe versus unripe tomatoes, outperforming YOLOv5. The robotic arm successfully manipulated tomatoes based on the vision model output. This system can rapidly and consistently sort tomatoes by ripeness level to aid packing and logistics operations. It demonstrates the integration of robotics and AI for agricultural automation. Benefits include increased efficiency, consistency, and meeting consumer demand over manual approaches.
Automated Tomato Maturity Classification: Integrating Computer Vision System and Robotic Manipulation
2024-07-17
1107203 byte
Conference paper
Electronic Resource
English
Research on Tomato Maturity Detection Based on Machine Vision
British Library Conference Proceedings | 2022
|Force and vision feedback for robotic manipulation of the microworld
Springer Verlag | 2008
|Bridging Language, Vision and Action: Multimodal VAEs in Robotic Manipulation Tasks
ArXiv | 2024
|