For Gord and Patrick Keller, 2019 was the last straw.
“Last year was that harvest from hell,” Patrick said, describing the frustration of trying to get a definitive grade on their canola in a year that saw frost in June, hail in July, then frost again in August.
“We drove samples all over the countryside and got a different answer from every pail — and not even remotely close.”
The Kellers farm 4,500 acres near Nut Mountain southeast of Kelvington, growing canola in rotation with wheat, barley, and peas. Gord has been at it for 47 years, while Patrick came home to join his dad 16 years ago.
Gord said canola grading hasn’t changed much over the years from the routine of crushing a sample over a perforated stick, then matching the colour of the crushed seeds by eye against an official paper card provided by the Canadian Grain Commission.
In a good year, Gord said the process is relatively straightforward: uniform yellow will most likely receive a number one grade.
In a less-than-ideal year, it gets complicated. For example, the CGC lists 23 factors to consider when grading canola.
The two most common downgrading factors are green seed and heated seed. Getting someone to make the call means bringing in a sample off the combine to the elevator.
“It takes me — our closest elevator is Wadena from Kelvington — it shoots the hell out of two hours for me to drive down there, get it rolled out and get back to the farm,” Gord said.
Surely, there had to be a better way. How about the computer right in your pocket?
A smartphone has processing power, a high-quality camera, and a light. Could it take a picture of crushed canola seeds in the field, compare it to other samples in a database and give farmers an answer on the spot?
In his previous career, Patrick was an instrumentation technician, writing software to automate gas plants. He knew it should be possible.
The father and son team got together with a developer to create Prograde. It’s taken a lot of work and about $125,000 so far, but the app was launched in September.
To use it, a farmer rolls out a sample of canola on site — Patrick used the hood of his truck this harvest season — then takes a picture of the sample. The picture is compared to samples in a database to deliver a verdict on expected grade. The process takes 10 to 15 seconds, and results are consistent.
“We feel like it’s 95 percent accurate, which in my opinion is 100 percent because it’s always the same,” Gord said, sharing that a neighbour had a load of canola rejected at one elevator then accepted as top grade at another.
While Prograde launched with 300 samples in its database, Patrick said the app is designed to add to this number as farmers use it. Using machine learning and artificial intelligence, it refines its accuracy the more it is used.
“It’s a great tool and I think that the technology definitely has the grounds for becoming a standard. There’s no doubt about it in my mind,” he said.
Getting to that standard means working with the CGC to arrive at some accepted grading factors. The Kellers are meeting with grain commission representatives to try to nail down some of the more subjective ones.
For example, “distinctly green” has an official colour on the CGC card, but Patrick said to add it to the app, they need an RGB value to program it into the software. Likewise, Prograde uses a smartphone’s onboard light to illuminate a sample, but how bright this light should be must be determined.
Another limitation of Prograde is that it tests 100 seeds, while CGC calls for 500-seed minimums. It’s also only available for iPhone, although the Kellers have been getting requests for an Android version.
Prograde is available at theharvestapp.com or Apple’s App Store.