AUGUST 2024
CELEBRATING 145 YEARS AS CANADA’S PREMIER HORTICULTURAL PUBLICATION
THEGROWER.ORG
THE RISE OF ROBOTICS
Artificial intelligence eyes auto-labour
Those headlights might look like two rising moons. But no, that’s a robot that’s moving from the twilight zone into reality. In this photo, a robot operating with artificial intelligence is in the second year of trials scouting PVY virus in Prince Edward Island potato fields. Photo by Charanpreet Singh.
Caution! Robots at work. That’s a safety sign that’s becoming more common in Canadian horticulture. A multi-disciplinary team of researchers at the University of Prince Edward Island (UPEI) is encouraged by a robot called AgriScout that’s using artificial intelligence to identify potatoes infected with potato virus Y (PVY). This groundbreaking, applied research is jointly funded by potato growers in New Brunswick and Prince Edward Island as well as their provincial governments. PVY causes potato disease which limits the yield and lowers the marketability of potatoes. In worst cases, it can cause necrosis in tubers. Symptoms are sometimes difficult to detect. Trained scouts must walk seed fields to identify and pull out plants that exhibit PVY symptoms – a practice called rogueing. “It’s tough work,” says Ryan Barrett, research and agronomy specialist with the PEI Potato Board. “It’s hard
to find trained people to rogue seed potatoes and it’s becoming harder to identify the newer PVY strains.” That’s why growers have turned to UPEI’s Precision Agriculture Research Group for a practical field solution that leverages artificial intelligence to efficiently detect PVY infections. “Farming is not new to us,” says Aitazaz Farooque, director of the Canadian Centre of Climate Change and Adaptation, UPEI. Aitazaz with his co-investigator -- Dr. Gurjit Randhawa, assistant professor, School of Computer Science, University of Guelph and their Master’s student Charanpreet Singh -- have been tweaking the AgriScout, building the image libraries, creating databases of symptoms, and adding a Starlink service to upload geo-tagged photos from remote areas. “One of our challenges was that a bright sunny day offers too much contrast to distinguish details of leaf shape and colour,” adds Randhawa. “We’ve added a canopy over the camera’s boom to address that issue and it is working quite well.”
Ryan Barrett was in test fields with the UPEI scientists in early July 2024. Now in the second year of field trials, a database of AI learning contains more than 80,000 labelled photos that are used to identify infected plants with more than 80 per cent accuracy. In one test in early July, the AgriScout surveyed a small field of potatoes known to be planted with about one per cent PVY, geo-tagging suspect plants with a real-time kinematic GPS and creating a map of PVY infections. After the potential infections were identified, the experienced roguers followed the map to verify the PVY infections. Out of 30 randomly flagged plants, 25 were positively identified as PVY, three were suspects and only two were false positives. That translates to an accuracy of 83 per cent. “We’re still in the learning process,” says Barrett, “and we still need to test the AgriScout with multiple varieties and growth stages. A future step, ideally, is to enable the robot to spray paint the infected plants so that anyone can pull them out.” Continued on page 3
Improving resilience of horticulture P6 5
EZ Grow Farms opens glasshouse PG 6
Equipment & farm machinery PG 12
Volume 74 Number 08 P.M. 40012319
@growernews
KAREN DAVIDSON
$3.00 CDN