Developments in computer vision and artificial intelligence are making some strategic changes to the way growers fight herbicide resistant weeds and offer the promise of significantly reduced herbicide use in the future. 

Drew Lyon, extension weed specialist at Washington State University, says the new site-specific weed management technology is currently popular in high-value crops, but steady engineering advances using software similar to that employed for facial recognition, along with more accurate on-the-boom sensors will soon offer growers of broad acre crops the same precision.

“John Deere is investing heavily in holdings related to machine vision and artificial intelligence, Bayer is developing a comprehensive weed identification photo catalog and Blue River Technology has fielded a sprayer with what it calls ‘see and spray’ technology,” he explains. 

As weed populations develop resistance to popular broad-spectrum herbicides such as glyphosate, the only alternatives are much more expensive chemicals, he says. “The only way to make those chemistries cost-efficient on a large scale, is to provide better sprayer technology to apply the herbicides only to targeted species, as compared with treating the whole crop with an over-the-top application,” he explains. 

That’s why firms like Blue River Technology are adopting the “every plant counts” mentality, and specifically engineering their equipment to identify weeds at various growth stages and apply herbicides only to those plants.

“Using computer vision and artificial intelligence, our machines can detect, identify and make management decisions on-the-go on every single plant in the field,” Ben Chostner, Blue River’s vice president of business development, explains. “This technology can be used to treat each plant precisely, whether it be for herbicides to the weeds, or fertilizer or fungicide to each plant that needs it.”

Blue River Technology’s literature estimates the use of this advanced engineering can eliminate up to 90% of herbicide use through precisely targeted application.

Similarly, Trimble’s WeedSeeker uses the well-known GreenSeeker infrared technology to apply herbicides directly to weeds and not to bare ground, a boon for weed control on fallow fields or row crops and pre-plant and post-harvest weed control operations.

“AI holds a lot of promise if it can be used to differentiate between weeds and cash crops…” – Drew Lyon

Lyon says in his area of the Pacific Northwest traditional crop farmers are showing interest in both WeedSeeker and WEEDit despite the costs of such systems. WEEDit, an optical day-or-night, pulse-width modulated spot spraying system searches for weeds on the go at 40,000 times per second and uses LED lights to signal its computer processor of the presence of weeds.

WEEDit recently introduced its Quadro system, which mounts monitors roughly every 10 inches of spray boom width —  up to 118 feet — using blue LED lights for improved weed identification day or night. Company officials say the system can be used for VRA of various farm inputs, weed biomass mapping and coordinating with application maps.

“Overall, I think we’re at a point with machine vision and artificial intelligence similar to where we were with GreenSeeker technology in the 1990s,” Lyon explains. “AI holds a lot of promise if it can be used to differentiate between weeds and cash crops, and I’m confident we’ll get there soon.”

The improved precision of even today’s products are making a dent in the way growers are fighting weed resistance, Lyon says.“Using drones, sprayer-based weed mapping and records kept by combine operators, we’re trying to drive weed numbers down. In doing so, we usually see weed concentrations becoming somewhat patchy in the field,” he explains. “U.S. and Australian researchers are working closely on this ‘divide and conquer’ strategy, because it will allow us to more efficiently treat weed concentrations.

“Our research has shown spraying systems such as WEEDit and WeedSeeker can greatly reduce costs of herbicide use, particularly during the later stages of summer fallow when weed cover is often below 50% and patchy.”

Lyon says as weed sequestration techniques improve, it will open up further development of the use of camera and computer-driven application systems.

“Already, here in the PNW, we have a big problem with rush skeletonweed, a perennial forb that reproduces by seed and through root sprouting,” he explains. “Our growers are using WEEDit to use chemicals which are much too expensive to apply over-the-top in its control. The improved spot spraying precision makes it economically feasible.”