A team of plant specialists is using artificial intelligence to grow healthier grapes. To that end, the researchers have outfitted robots with computer vision technology that can identify plants that have been infected with powdery mildew spores.
The result is essentially a modified version of facial recognition that has been trained to recognize mildew instead of human faces. Credit for the concept goes to Lance Cadle-Davidson, a biologist with the School of Integrative Plant Science (SIPS) who is currently trying to develop grape varieties that are resistant to the destructive mildew fungus. He needed a way to analyze and classify a large number of leaves in a short amount of time, and AI analysis had the potential to be far more efficient than manual examination.
With that in mind, Cadle-Davidson worked with the Department of Agriculture’s Agricultural Research Service (USDA-ARS) and the Grape Genetics Research Unit in Geneva, New York, to build camera robots (dubbed BlackBirds) that could scan leaves more quickly than the human eye. SIPS engineer Yu Jiang built the classification software, using feedback from Cadle-Davidson’s team to improve the accuracy of the AI system.
In doing so, the researchers essentially automated the leaf inspection process. The robots’ high-throughput phenotyping has significantly accelerated Cadle-Davidson’s research, to the point that experiments that used to take six months can now be completed in as little as a day.
The BlackBirds themselves are able to examine leaves at a microscopic level, gathering information at a scale of 1.2 micrometers per pixel. Cadle-Davidson is looking specifically at grape leaves, but believes that the system could be trained to look at other diseases in other plants, or modified to do the same for animals or the human medical sector.
In that regard, the BlackBirds have already drawn considerable scientific attention. The researchers have received a $100,000 grant to provide BlackBirds for other crops, and a $150,000 grant to upgrade the robots with infrared capabilities. Their article also picked up the 2021 best paper award from the American Society of Agricultural and Biological Engineers.
This is not the first time that someone has tried to use artificial intelligence to improve agricultural outcomes. ZenaPay has previously outfitted drones with software to track the growth of cannabis plants, while start-ups in India are using facial recognition and biometric wearables to make the dairy industry more efficient.
August 17, 2021 – by Eric Weiss