Cropify offers growers technology for objective grain classification

Andrew Hannon and Anna Falkiner of Cropify. “The grade can be misclassified and that is what Cropify wants to get rid of.” - Photo: Cropify
Andrew Hannon and Anna Falkiner of Cropify. “The grade can be misclassified and that is what Cropify wants to get rid of.” - Photo: Cropify

Growers will soon be able to use artificial intelligence for grain classification on the farm. A new South Australian technology can replace the current subjective assessments, says co-founder Andrew Hannon of Cropify.

“Cropify will allow growers to make a more informed decision about their grain early in the supply chain”, Mr Hannon emphasises. “Whether they will store their grain on-farm, sell in the domestic market or put it into the export market.”

Covering the whole supply chain

The technology is designed to be a solution for growers right through to bulk handlers, marketers and exporters, covering the whole supply chain. “For everyone there is an opportunity to have a rapid, repeatable, consistently accurate solution, to replace the current subjective assessments”, Mr Hannon says.

Mr Hannon comes from a farming background and has experience in grain farming as well. For the past 20 years he was involved in the grain supply chain, as a commercial manager of a major bulk handler in Australia.

Need for an objective grain classification system

After working in the industry, he understood there is a need for an objective grain classification system. “For a long time, we have been looking at a solution for the subjective nature of grain classification”, he points out, “In Australia, when multiple classifiers look at the same samples, they come up with different results. Obviously, if we can get a machine to do that job, it can do the job more accurately and more repeatable than a person across the whole supply chain.”

The same grain sample will always get the same result

Currently, the classification results of the grain derive the grade. “And this grade has a value”, Mr Hannon says. “But that grade can be misclassified and that is what Cropify wants to get rid of. The same grain sample will always get the same result. Now it’s all done by eye. With people doing this, you sometimes get varied results.”

Repeatability and accuracy

With their Cropify technology, Mr Hannon and his partner Anna Falkiner say they have created a uniform pulse classification system. The technology, based on artificial intelligence, should enable growers, grain handlers and marketers to classify a pulse sample in a way that is objective and repeatable. “Even though it’s early, we think we that with Cropify, we have been able to create that repeatability and accuracy.”

Mr Hannon says grain classification is currently time consuming. “Particularly with pulses. There are a lot of subjective tests involved in pulse classification, and a classification can take up to 20 minutes at the point of receival. Cropify will run a test with an immediate result. We are initially working on lentils and plan to expand first into other pulses.” De technology will also provide information on contaminants, variety and geolocation data.

AI technology in the horticulture industry

Mr Hannon was working for an agribusiness, when he noticed the AI technology being used in the horticulture industry. “I was consulting to the Department of Agriculture, looking at all the different types of technology available, and how that technology relates to quality control for plants and plant products, being prepared to export. I noticed there already was a solution out there for horticulture. I could see that a similar solution could be used for grains and pulses.”

With a scientist that had worked on this concept, an algorithm for pulses was built. “We used the algorithm and machine learning to identify some of the major defects in lentils. But we will expand that into other commodities.” The company also works with seed breeders, to make sure the technology will be able to recognise new varieties.

Users that have a Cropify-subscription, will be able to use an image of a grain sample to analyse it online. “We have a hardware solution that we have designed and built here, that will enable users to present a sample to the camera”, Mr Hannon says.

The Cropify algorithm will identify whether the sample is sound or defective. If the sample is defective, Cropify will show what is defective for. “That can be poor colour or insect damage for example”, Mr Hannon explains. “Is it chipped or broken? All the major defects the grain industry classifies for.”

We see a trend towards more and more on-farm storage

Mr Hannon says that Cropify should be accessible up and down the supply chain. “Growers in the paddock should also be able to use it. We see a trend towards more and more on-farm storage. Those growers, that store grain on the farm, want to make a decision as to where they deliver their grain. But we are also building it to meet the bulk handler market and to make Cropify a major classification tool, both here in Australia and overseas.”

Text continues below image

Initially the technology is to be used for small red lentils. Later it also will be suitable for other pulses and coarse grains. - Photo: Lukasz Rawa
Initially the technology is to be used for small red lentils. Later it also will be suitable for other pulses and coarse grains. - Photo: Lukasz Rawa

Small red lentils

Initially the technology is to be used for small red lentils. Later it also will be suitable for other pulses and coarse grains. Small red lentils are the hardest to classify in Australia, Mr Hannon says. “We thought we would start with the most difficult. If we can get a good solution for small lentils, then a solution for larger commodities like chickpeas and fava beans will be straight forward. We will also move on to coarse grain and in the future possibly to other food products. The potential is huge.”

The results of the technology have been tested by major Australian independent laboratories. “This way we can prove the concept and confirm the results with gold class classifiers in the laboratory. But one of the great things about machine learning is that, each time with a new image, the image library gets bigger and the algorithm gets smarter.”

So far, there is a lot of interest right across the supply chain, Mr Hannon says. “That’s why we are moving rapidly to get to the commercials trials as fast as we can.”

The South Australian government recently awarded Cropify a grant as part of its AgTech Growth Fund. “This will help to get to the minimum viable product stage”, Mr Hannon explains. “Then we will move through a series of commercial trials and into commercialisation. We expect the commercial trials to start over the coming months.”

Groeneveld
René Groeneveld Correspondent for Australia



Beheer