Agriculture is one of the most ancient human activities, which will never lose its value due to mankind’s need for nutrients. Agricultural methods certainly change over time with old technologies being replaced by new ones.
As manual harvesting of fruit and vegetables is quite expensive and time consuming, farmers around the world have been trying to automate the process. The main difficulty is to teach the machine to distinguish ripe fruit and vegetables from unripe ones.
Vegebot is a device for harvesting iceberg lettuce. To teach the machine to identify ripe lettuce, scientists applied machine learning. The robot has studied a great number of images, 3D-models and even real lettuce heads both under lab and field conditions, whereupon the accuracy with which it identifies ripe lettuce heads has reached 91%. It is achieved using the camera built into the machine.
The harvesting of lettuce is the responsibility of the second module equipped with its own camera and knife, which were also put through machine learning. The percentage of lettuce heads prematurely cut by it is 38%, which is quite a lot. The whole process takes about 32 seconds. The machine is currently being tested on real garden beds in the UK.