A Computer Vision Based Robotic Harvesting System for Lettuce

Abstract

Harvesting lettuces is a challenging task in agriculture. In the paper, we propose a computer vision based robotic harvesting system for lettuce in hydroponic farms. In these farms, lettuces are grown in holes evenly located on parallel plastic tubes. Therefore, the problem of estimating locations of lettuces’ stems can be simplified by estimating the hole locations instead. In particular, locations of holes covered by lettuces are estimated based on uncovered holes and/or tubes’ edges which are found by using Hough transforms. A robot manipulator and a servo controlled gripper are used for the harvesting task. Experiment results show that the system, with the mean location error of 0.83 mm (typical cases) and maximum error of 9.62 mm, can efficiently perform harvesting task in real word environments.