快猫短视频

Artificial intelligence is getting down in the weeds. An AI-powered robot that can distinguish weeds from crops and remove them could eventually be used as an alternative to chemical herbicides.

Kevin Patel and Nihar Chaniyara at tech start-up AutoRoboCulture in Gandhinagar, India, have created a prototype device, called Nindamani, specifically for cauliflower crops.

The is powered by a pre-existing . The duo trained the AI using more than six hundred images of cauliflower crops, in which the crops and weeds in each image were manually labelled.

The AI can visually process an area of 0.2 square metres every second, and has learned to distinguish between young cauliflower crops and weeds with an accuracy of about 85 per cent, says Chaniyara.

Nindamani is mounted on wheels and equipped with a camera and robotic claw. When it detects a weed, the robot removes it using the claw and discards it into the furrow between rows of crops.

It is battery-powered and can work for up to eight hours, removing weeds at a rate of one every five seconds. Nindamani isn鈥檛 fully autonomous yet and must be wheeled along a row of crops.

Another current limitation is that the robot can only remove relatively small weeds up to around 30 centimetres, says Patel.

Patel and Chaniyara both have relatives who are farmers. 鈥淚n India, there is a labour shortage problem, so they use chemical sprays to kill weeds,鈥� says Chaniyara. He says they hope this could eliminate some use of pesticides.

Though this prototype was used on cauliflower plants, the duo say they have also successfully tested the robot on tomato, aubergine and okra crops. They plan to expand to cotton and peanuts next, and also to equip Nindamani with or so that it can move autonomously within the boundary of a farm.