RED DEER, Alta. — A Saskatchewan company plans to use artificial intelligence and machine learning to give farmers and grain buyers an edge by boosting the accuracy and speed of how grain is graded.
Ground Truth Ag is developing systems that could be used in places ranging from elevators to farms, including grain bins and combines as well as precision agriculture, said chief operating officer Divyesh Patel.
It promises to revolutionize a process for grading grain such as wheat that largely hasn’t changed in decades, he said.
“It’s a monumental change. I mean, even today, most of the visual characteristics are graded by hand and graded by eye.”
The problem is that each person tends to differ in their assessments, said Patel.
“If you were to grade some grain and I was to grade some grain, we might look at it and see something a little bit different, and that causes friction in the transaction between farmers and buyers.”
He said his company has created technology that will not only consistently give the same results with greater accuracy and transparency, but also assess samples more quickly.
“It’s much faster than a human grader; on average during harvest, it might take about nine to 10 minutes to create a sample of grain. Our system can do it in less than two minutes.”
The Regina-based company was honoured with a Farmers Choice Award of $5,000 Nov. 9 at the Ag Innovations Awards Competition held as part of the Agri-Trade Equipment Expo in Red Deer.
Finalists made their pitches to a panel of three judges. An audience of more than 150 farmers also voted via their smartphones on whether each company solved an urgent or significant challenge faced by producers.
Billion-dollar investments by internet giants such as Alphabet Inc., parent company of Google, in research on artificial intelligence and machine learning have lowered the bar for early start-up companies such as Ground Truth Ag, said Patel.
“I think if you were to try to have done what we’re doing (now) five years ago, it wouldn’t have been possible. The cost to do what we’re doing would have been astronomical, and you couldn’t raise enough money to do it.”
One of the systems that Ground Truth Ag is developing is based on computer vision. A camera takes pictures of grain samples using bright white light, he said.
“It’s just a camera, but the real magic is in the computer vision models. Those images are fed to our machine learning system. It looks at those pictures, it understands what’s in those pictures, and then we build a prediction model based on those images.”
The model can then be applied to new sets of images, he said.
“The technology isn’t in the camera itself; it’s in the software that looks at those pictures and understands what’s in them.”
Besides identifying each kernel, it can also examine whether each one is healthy, “and if it’s not healthy, we identify what is afflicting that kernel. We identify impurities — ergot, other grains, whatever might be in the sample.”
The company is currently targeting broad acre farming as its niche market, said Patel. The first type of grain will be hard red spring wheat, followed by Canada western amber durum and crops such as red lentils and peas, he added.
However, the technology could eventually encompass other types of crops around the world, he said.
“I mean, the sky’s the limit, right? … Anything that’s being visually graded can be done.”
Ground Truth Ag has developed bench-top units that will use camera vision to analyze the visual characteristics of hard red spring wheat. The company expects to start delivery of the units by June or July at a price of about $15,000, said Patel.
However, a discounted price of $10,000 has already been made available to the first few committed early customers, he added. The company will also be charging an annual access fee of about $1,500 for upgrades, he added.
Future plans for the desktop units include fitting them with near-infrared spectroscopy, which will allow them to also examine non-visual characteristics such as protein and moisture levels, said Patel. Users such as grain elevators and facilities that contain laboratories are a natural fit for the units because that’s where most of the grading is currently taking place, he said.
Although Ground Truth Ag didn’t start out with the goal of reducing friction between farmers and grain buyers, “it is a pretty important part of the product now,” he added.
“And I don’t think anybody’s doing anything untoward or nefarious. I think it’s just that there is imprecision in the grading process, and this computer vision can remove that imprecision and create more transparency in the process.”
The bench-top units are currently being evaluated by a commercial grain buyer, he said.
“They have it in their facility, and they’re running samples through it to see if it works for them. We’ve also talked to some inspection and certification companies who are interested in our product.”
However, producers could also use the technology directly on the farm.
“You can sample more frequently if you’re moving your grain from one bin to another, but you’re not ready to sell it yet,” said Patel.
“You can find out what you have in your inventory. Most people do the sampling when they load their bin, and you can assess it on your farm and then you can, as you’re unloading it, do another sample and integrate it as well.”
Farmers could be provided with invaluable knowledge about what’s in their trucks when they talk with buyers about what prices they should be getting for their grain, said Patel. Producers could also use the information to improve how they blend their grain on the farm before they sell it, he added.
Grain is priced based on its grade, with farmers earning premiums for protein content.
“Could you blend something that’s barely a No. 2 and mix it with your No. 1 grain on the farm and get a No. 1 price when you take it to the buyer?” said Patel.
Ground Truth Ag has also identified opportunities to use the technology in applications involving precision agriculture, he added.
“If you had location-based information about your grain quality, you could potentially change your agronomic practices in advance of fusarium damage appearing in certain parts of your field. Maybe you can do different rates of fertilizer or apply fungicide as appropriate in that field.”
The company is also developing a version of its technology that will be fitted on combines, which it expects to launch in 2025, said Patel.
Although some of the components for Ground Truth Ag’s systems, including cameras and computers, are from countries such as Germany, much of the manufacturing and assembly will be conducted in Saskatchewan, he added.
SASKTODAY.ca is Saskatchewan's home page. Bookmark us at this link.
