Research into accurately estimating total biomass of standing cover crops from a moving tractor may eventually spawn new ways for strip-tillers to save money on planting rates and herbicide use in row crops. 

North Carolina State University Extension weed specialist Ramon Leon says a study of “structure-from-motion” involving researchers from NCSU, Iowa State University, Texas A&M and the USDA’s Agricultural Research Service has shown accurate assessments of cereal rye cover crop stands can be made using inexpensive GoPro cameras and sophisticated computer analysis.

Cereal rye was used in the multi-year study, begun in 2021, because it is widely used across the U.S. and can suppress more than 90% of weeds under ideal conditions, primarily by creating a thick mulch. The study involved 5 fields at a research farm near Goldsboro, N.C., where cereal rye growth had shown significant variability over the years. The cover was planted in January 2023 with a drill, then terminated 4 months later with glyphosate and a roller-crimper.

Structure-from-motion (SfM) is the photogrammetric process of estimating the 3-dimensional structure of a scene from a set of 2-dimensional images. The 3D images are analyzed with computer algorithms and produce a synthetic visual estimate of total biomass in the scene.

Leon says the NRCS-funded study involved hand-held GoPro cameras walked through standing cover crops with resulting images analyzed by NCSU’s super computers. The results were used to accurately estimate the effectiveness of cover crops as agents of weed suppression. 

More Realistic

To add additional accuracy to the results, researchers are now using multi-lens cameras that provide a stereo image used to eliminate false estimates from tall, but very thin, stems found in the small-grain canopy.

“The use of 3D images provides a more realistic and accurate quantification of density, growth and biomass of the cover crop than would 2D images,” Leon says. “This is particularly useful since cover crop growth can be patchy in fields, making biomass estimation difficult.”

Since cover crop production helps blunt weed growth, SfM images and related analysis can be used to predict weed suppression levels in different areas of a field. Areas with the lowest biomass estimates may be the most likely to have late-season weed escapes.

“While the cover crop is growing, growers are deciding whether it will help with weed suppression,” Leon says. “If the cover did well in some areas but not others, they may divide the field into sections and manage weeds in each section differently. If the cover is too sparce or small, the farmer may decide to terminate it early and incorporate it to reap organic matter benefits — foregoing any weed control potential. In the latter case, the decision might be to use a specific pre-emergence herbicide program or plan to take earlier-post-emergence action.”

POWER POINT. Cameras are mounted ahead of the tractor to record video over the cover crop immediately ahead of termination. April Dobbs

Leon says with potential weed probabilities known ahead of the cash crop growing season, growers can better plan field visits to scout areas more likely to have weed escapes. And at planting, they might decide to increase seeding rates only in susceptible areas to improve in-season crop canopy coverage, rather than using a single heavier rate across the entire field. Both practices can protect yields and save time and money. 

Similarly, in the case of developing corn, the technology could be used to measure the corn’s canopy potential at V4 while simultaneously estimating weed biomass between rows — to better manage herbicide and nitrogen (N) use, he notes. 

“There is talk among my colleagues about further developments of this system, with the addition of artificial intelligence, to measure early-season ear count and in the case of cotton, estimate boll size,” Leon says. “Those applications are still in the more distant future, but 5 years ago I thought the project we’re now discussing might take 25 years to develop to its current state.”

Doing What’s Practical

April Dobbs, a PhD candidate with Leon during the NCSU GoPro study and now a research development specialist with BASF, says manually assessing field-scale cover crop biomass is impractical.

“Manually harvesting and weighing samples is a laborious, time-sensitive process that is not practical on field scale,” she says. “Also, satellite and UAV-based remote sensing methods lose their sensitivity when the cover crop canopy closes, and they don’t capture the variability of individual fields and growing conditions. 

“Current drone technology really doesn’t provide a very accurate estimate of that variability in biomass either,” she says, noting UAV measurements favor canopy height, and low-flying drones disturb the canopy with prop wash, adding variability not found in hand-held or tractor- or sprayer-mounted cameras.

Comparing the project’s image-based estimates and field samples taken manually, she said the new technology was highly accurate.

Proof of Concept

Leon says the project proved the technology is viable but notes several challenges must be overcome before farmers see it introduced commercially.

“First, we know satellites and drones aren’t currently acceptable remote sensing platforms for 3D assessment of cover crop biomass,” he says. “So, to save growers the cost of a separate field trip to generate the images, we propose mounting the cameras on the front of a tractor or sprayer used to terminate the cover crop.”

Since the cameras need to be very stable during their imaging, that presents an engineering problem of overcoming vibration and jostling of the vehicle, with some sort of shock-absorbing mounting fixture to compensate for unwanted motion. 

“That’s something we’re contemplating right now, but it’s a problem we need to solve before a grower can generate the 3D images during the termination pass,” Leon says.

Also, computing power is a limiting factor if one envisions an on-board tractor-mounted 3D image system.

“We’re dealing with billions of data points with this technology and we’re currently using very powerful computers based on campus to analyze our research study samples,” Leon says. “With current technology, on-board systems just don’t seem practical for real-time results.”

How Strip-Tillers Could Benefit

He envisions the technology becoming available with farmer-produced images uploaded to the cloud and processed as a service with minimal turnaround intervals. 

“Commercial fishermen are using this kind of system to generate images from the ocean with sonar and tracking potential catches, with results returned to them within hours to days,” he says.

Another approach might be results based on various levels of detail, or layering, Leon says.

“A grower might say I’m seeing the density of my cover crops as I’m spraying, but I might need a reminder of that in map form,” he says. “Current stereo camera technology is sufficient to provide that data with a very fast turnaround for usable maps.

“So, depending on where you are and availability and speed of your internet service, you might be able to get reports in different modules or levels of complexity, like those tax software programs where you get the free basic plan, but plans with higher levels of detail will cost you more.”

Leon says the technology will help conservation-minded farmers like strip-tillers make sound management decisions to improve weed control in a cost-effective manner, a factor that will help them stick with the practice instead of returning to full tillage.

“We want to maintain no-till or reduced tillage as much as possible,” he explains. “And, if herbicides don’t work, then growers are more likely to return to cultivation and soil disturbance.”