LETTUCE is California’s main vegetable crop. The state grew $1.6 billion-worth of the leafy plant in 2010 and accounts for more than 70% of all lettuce grown in America—itself the world’s second-biggest exporter of the stuff. It is a fiddly business. As well as having to be fertilised and weeded, lettuce must also be “thinned” so that good plants do not grow too close to each other, inhibiting growth. Much of this is still done by hand. Labourers, who tend to be paid per acre, not per hour, have little incentive to pay close attention to what they pull from the ground, often leading to unnecessary waste.
Enter Lettuce Bot, the brainchild of two Stanford-trained engineers, Jorge Heraud and Lee Redden. Their diligent robotic labourer, pulled behind a tractor, starts by taking pictures of passing plants. Computer-vision algorithms devised by Mr Redden compare these to a database of more than a million images, taken from different angles against different backdrops of soil and other plants, that he and Mr Heraud have amassed from their visits to lettuce farms. A simple shield blocks out the Californian sun to prevent odd shading from confounding the software.