Automated Tomato Inspection and Harvesting System Using Robotic Arm and Computer Vision in Greenhouses

Abstract

This paper discusses the development of an advanced robotic system equipped with a computer vision system for use in greenhouses to identify and harvest ripe, unripe, and diseased tomatoes. The system operates in two modes, harvesting and pruning, and is implemented using a raspberry pi and free software libraries such as Python, OpenCV, and PyTorch. The vision system utilizes YOLOv3 deep neural network for classification and MiDaS for depth estimation, and a graphical user interface is included for communication with the farmer.