Saskatchewan company working to automate grain grading in real-time


A Saskatchewan based start-up is deploying a combination of cameras and machine learning to automate and dramatically reduce the time it takes to grade a sample of grain.

Ground Truth Ag has developed a machine with a vision detection model that can assess around 50 different grading factors for Canada Western red spring wheat, as defined by the Canadian Grain Commission, in near real-time.

“We’re starting with the benchtop version, but eventually we’ll have this unit on combines and potentially on augers throughout the supply chain,” explains chief operating officer Divyesh Patel in the interview below, filmed at Ag Days in Brandon, Man. “I would love to see it every place where people want to understand the grade of grain.”

Initially, the company anticipates selling the benchtop units to grain elevators that process hundreds of samples each day.

“We can do it in two minutes, as opposed to a human grader that might take nine or 10 minutes,” says Patel, as he demonstrates how the unit works. “I don’t know that it will necessarily replace graders, but it will just augment their skills. They won’t have to look for all of the factors — they can look for the ones that the unit doesn’t do as well, and focus on those instead of ones that we can sort of tear out in a minute. We see it as a tool, not necessarily a replacement.”

On farms, he sees a use case in situations where grain handling equipment is capable of up-blending to meet a higher grade.

The company is planning to conduct another year of field trials with its more-rugged combine-mounted version. In addition to providing a more representative picture of quality than a traditional grading sample, the goal is to use real-time grain quality data from the machine to inform agronomic decisions, such as top-dressing fertilizer and creating variable rate prescriptions based on variations in grain quality across a field.

“We did a little study this year during harvest where we harvested a small field plot, and we found significant variation, even within 20 yards, the protein concentration would go from 17% to 12%,” notes Patel.

Check out the video above for more with Divyesh Patel on Ground Truth Ag’s effort to advance automated, real-time grain grading.

Related: Saskatchewan startup secures $4 million to develop grain analysis tech

Wake up with RealAgriculture

Subscribe to our daily newsletters to keep you up-to-date with our latest coverage every morning.

Wake up with RealAgriculture

Please register to read and comment.


Register for a RealAgriculture account to manage your Shortcut menu instead of the default.