Drone operators explore machine learning

Indeed, artificial intelligence, machine learning, and a better understanding of scalability are giving drone operators the capability to gather information at the millimetre level – and interpret it for farm clients in a practical way.

Results, not just pretty pictures

Norm Lamothe, a grain farmer from Ontario, Canada, and unmanned aerial vehicle (UAV) operator, has been using drones on his farm – and for farm clients – for years. In his experience, the usefulness and profitability of drone tech has been rising as data-building and layering techniques have improved. This wasn’t necessarily the case 5 years ago.

Also co-founder of Deveron UAS – a drone and data company operating in both Canada and the United States – Lamothe says Silicon Valley promoters and data companies initially promised many things that drones could not achieve at the time. Drones were just creating siloed field images and “static data” that farmers could not use.

“[Drone tech promoters] didn’t have a lot of know how about grower challenges,” says Lamothe. “You really have to listen to growers and what they can do at the field level […] Growers don’t have room to take risks.”

_{Text continues underneath image}

The usefulness and profitability of drone tech has been rising as data-building and layering techniques have improved. – Photo: Galama Media

Low-hanging fruit

The “low-hanging fruit” is what drone and ag-data companies are looking for now. Lamothe says this includes services focused on variable rate scripting, imagery for variable rate nitrogen and fungicide applications, and those combining mid-season drone flights (to analyse visible crop health) with soil sampling to highlight potential nutrient deficiencies.

“What I see on my farm and the decisions we made, there’s opportunity to extract another 200 to 300 dollars per acre,” Lamothe says. “We don’t farm a lot of acres so we have to farm really well. It’s farming to the acre now as opposed to farming the field.”

Population and predictive nutrient mapping

Mike Wilson, program lead with Veritas Farm Management – a Canadian ag-service company and subsidiary of Deveron UAS – says his company has had success using drones to address plant population variability. This involves flying a small drone close to the ground shortly after crop emergence. Key checkpoints – determined based on zone maps – are photographed, and stand gaps, even very small ones, are identified. The client can then replant those areas either manually or using variable rate equipment.

An entire field can be accurately covered in minutes, which Wilson says makes it both more time efficient and thorough than physically walking the field. The same stand information could also be used to make crop insurance claims.

_{Text continues underneath image}

Lamothe says the next step in drone tech involves AI and machine learning – or developing intelligent machines that can make independent, clearly communicable in-field decisions. – Photo: Bert Jansen

Predict nutrient availability

Wilson and his colleagues are also researching how drones can be used to predict nutrient availability – for crops the following spring – by looking at how biomass and crop colour (leaf data) in red clover can indicate accessible nitrogen.

Lamothe adds the ability of drones to generate detailed and consistent data sets also makes them an important tool in the general research sense.

_{Ontario grain farmer Norm Lamothe:}

It’s a snapshot in history. That’s where the high resolution and frequency of collection matters

“Drone data doesn’t lie,” says Lamothe. “Anything that’s research based has a strong case for drones because you have a pretty strong layer of static data that you can review […] It’s a snapshot in history. That’s where the high resolution and frequency of collection matters.”

Machine learning the next hurdle

Lamothe says the next step in drone tech involves AI and machine learning – or developing intelligent machines that can make independent, clearly communicable in-field decisions. Steve Laevens, UAV operator and sampling coordinator with Veritas, says this technology is already being used by his company in its stand-count service. The same principal also applies to other agronomic measurements, from identifying manganese deficiency and weed pressure to finding rocks, trash, or other obstructions that could damage equipment.

_{Text continues underneath image}

The accuracy and reliability of satellites has come a long way, and they provide good information at the metre level. However, drones can operate on a millimetre scale and don’t have to worry about barriers like cloud cover. – Photo: Ruud Ploeg

“We fly the drone and tell it ‘that’s a corn plant.’ It keeps going, sees something else green, and asks ‘is this a corn plant?’ We say no. It keeps going and asks [again] and we say yes. Do that enough times and it learns,” says Laevens. “That becomes machine learning as the data grows.”

Wilson says Veritas’ current drone service costs around $ 2.00 to $ 3.00 per acre for the flight, and another $ 2.00 to $ 3.00 for basic analysis. He and Laevens also say the expense of more extensive analysis should drop as machine learning capabilities increase.

Satellites vs. drones

Lamothe says limitations still exist for drones when it comes to scalability. Satellite imagery, for example, is a complimentary tool that can provide cheaper options for covering large acreages. Satellite data, however, is not as detailed.

Wilson expresses a similar sentiment. The accuracy and reliability of satellites has come a long way, he says, and they provide good information at the metre level. However, drones can operate on a millimetre scale and don’t have to worry about barriers like cloud cover.

“Drones will always do things satellites don’t. They are on different platforms,” he says. “Satellites are just as important. They are probably more scalable.”

_{Text continues underneath image}

“We fly the drone and tell it ‘that’s a corn plant.’ It keeps going, sees something else green, and asks ‘is this a corn plant?’ We say no. It keeps going and asks [again] and we say yes. Do that enough times and it learns,” says Laevens. – Photo: Henk Riswick

Agronomy basics always important

Wilson reiterates drone tech, as the technology currently stands, is just one of many strategic tools. “I’m a firm believer in soil sampling. Is still very valuable right now […] I think it would be nice to do everything from the air, but I still think a collaborative approach holds value. That includes soil sampling and getting in the field with an agronomist.”

Lamothe adds the use of drone technology is still comparatively new to many farmers, and that widespread adoption of new technologies takes a while to pick up.

“That adoption curve for a product that everyone uses now is an example of where every new technology in the marketplace wants to be,” says Lamothe. “It’s got to be so simple to use and cost effective that everyone wants to use it every year. That’s what we’re trying to do.”

Also read: Agricultural drone market to grow to USD 4.8 billion