Robotic Harvesting of Asparagus using Machine Learning and Time-of-Flight Imaging – Overview of Development and Field Trials

Abstract

Asparagus is a problematic crop because it grows so quickly that it requires harvesting every one or two days. Asparagus farms require typically 8 workers per hectare for harvesting during peak season. The subsequent labor issues this causes, means it is an ideal crop for robotic harvesting. Several prototype robotic machines have been trialed with limited success. This work investigates the combination of Machine Learning and Time-of-Flight cameras to locate the cutting point of the asparagus spear when travelling at a constant speed of 0. 33m/s. A ‘proof of concept’ machine was developed to validate the detection system and demonstrate harvesting with a rudimentary robotic arm and end effector. Field trials showed the arm harvested 92.3– of the targeted spears. However, it was found that multiple arms will be required to be commercially viable.