How Machine Vision Helps Automate Lettuce Harvesting
Lettuce is a valuable crop in Europe and the US, but a shortage of workers makes harvesting this field vegetable difficult, since finding enough seasonal labor to meet harvest commitments is one of the biggest challenges in the sector.
Moreover, as wage inflation outpaces producer prices, margins are very thin. In England, agritech specialists and equipment makers are working with IDS Imaging Development Systems GmbH, based in Obersulm (Germany), to develop a robotic solution for automating lettuce harvesting.
The team is working on a project funded by Innovate UK and includes experts from the Grimme agricultural machinery plant, the Agri-EPI Centre in Edinburgh (UK), Harper Adams University in Newport (UK), the Centre for Machine Vision at the University of the West of England in Bristol, and two of the largest lettuce producers in the UK, G's Fresh and PDM Produce.
As part of the project, existing leek harvesting equipment is being adapted to lift lettuce from the ground and grip it between pressure belts. The outer or "wrapper" leaves will be mechanically removed to expose the stem. Machine vision and artificial intelligence are then used to determine the exact cutting point on the stem to neatly separate the lettuce head.
“According to colleagues at Salad Harvesting Services Ltd., a G subsidiary, cutting the iceberg is the most technically challenging stage of the automation process. The robot harvester prototype being developed includes a GigE Vision camera from the uEye FA family. It is considered especially robust and therefore well suited to demanding conditions. Since this is an outdoor application, a housing with IP65/67 protection is required”
The choice fell on the GV-5280FA-C-HQ model with a compact Sony IMX264 CMOS sensor with global shutter.
“The sensor was chosen mainly for its versatility. We don't need full resolution for AI processing, so sensitivity can be increased through binning. A larger sensor format also means wide-angle optics are not needed”
The robot lettuce harvester prototype will be used for field trials in England near the end of the season.
“We are glad to be part of the project and look forward to the results. We are convinced of its potential to automate and improve lettuce harvesting efficiency, not only in terms of compensating for the shortage of seasonal workers”
The challenges facing the agricultural sector are truly complex. According to the Food and Agriculture Organization of the United Nations (FAO), agricultural productivity must grow by almost 50% by 2050 compared with 2012 due to a sharp rise in population.
This expected yield growth poses a huge challenge for the agricultural industry, which is still in its infancy in digitalization compared with other sectors and is already under strong pressure to innovate due to climate change and labor shortages.
The farming of the future is based on networked devices and automation. Cameras are an important building block, and artificial intelligence is the central technology. Intelligent applications such as robot harvesters can make a significant contribution.
In summer 2021, it became known that engineers had developed a vegetable-harvesting robot that uses machine learning to identify and pick not only common but also complex agricultural crops.
Vegebot, developed by a team at the University of Cambridge, was initially trained to recognize and harvest iceberg lettuce in laboratory conditions. To date, it has successfully passed tests in various field conditions in collaboration with G's Growers, a local fruit and vegetable cooperative.
Although the prototype is far from as fast and efficient as a human, it demonstrates how robotics can be expanded in agriculture, even for crops like iceberg lettuce that are particularly difficult to harvest mechanically.
Crops such as potatoes and wheat have been harvested mechanically for decades, but many other crops have resisted automation. Iceberg lettuce is one of them. Although it is the most common salad variety grown in the UK, iceberg is easily damaged and grows relatively flat against the ground, which poses a problem for harvesting robots.
"Every field is different, every lettuce is different. But if we can get a harvesting robot to work with iceberg lettuce, we can make it work with many other crops," said Simon Birrell of the Cambridge Engineering Department.
Vegebot first identifies the "target" crop in its field of view, then determines whether a particular lettuce is healthy and ready to harvest, and finally cuts the lettuce from the rest of the plant without crushing it so it is "supermarket ready". For a human, the whole process takes a couple of seconds, but for a robot it is a really complex task.
