Researchers using high-tech cameras, machine learning to answer questions

Oftentimes, farmers provide ideas for research that can improve quality and yield of their crops. Three LSU AgCenter scientists who work continuously with emerging technologies — Tri Setiyono, Randy Price and Kevin Hoffseth — are each currently studying crop production and post-production projects that directly respond to requests from farmers:

Tri Setiyono, an assistant professor, has been studying how extreme weather events like heavy rain and flooding effect nitrogen fertilizer effectiveness in corn and establishment of stands in soybean. Producers wanted help determining mitigation options — whether they should, in these conditions, supplement nitrogen fertilizer in corn and replant soybean.

Randy Price, an engineer, is researching different drone nozzle types and their impact on crop infiltration, as well as researching crop coverage at different spraying heights. These trials are in response to requests from farmers that spray with drones.

Kevin Hoffseth, an assistant professor, is developing an automated system to more efficiently determine the quality of harvested soybeans. Farmers had approached the AgCenter for ways to help ensure they weren’t missing out on potential value for their grain loads.

Hoffseth submitted a full patent application for a high-tech system of digital image acquisition and data analysis of soybeans. He and his graduate students designed this configuration to run numerical calculations that consistently and quickly analyze the quality of soybeans based upon color, shape and texture.

“This technology is designed to help the farmer quickly know what sort of damage there might be and help him decide whether to sell the beans or not,” said Hoffseth. “We’re developing ways to improve and speed up the processing of large amounts of complex data.”

Hoffseth and his students recently purchased a highly specialized camera which can compare images taken in the visible spectrum to images in the shortwave infrared spectrum.

“That camera should give us valuable data on the oil and protein content in the soybean,” said Hoffseth.

These developing technologies could provide a more accurate and consistent way of understanding the quality of grain in the field or in the warehouse.

“It can benefit both the buyer and the seller, saving them time and improving accuracy,” said Hoffseth.

Tri Setiyono looked at simulated flood conditions for soybean and corn on different soil types.

“The concern after flooding is, if it causes leaching in N fertilizer, how much fertilizer should be put back to avoid significant yield loss,” said Setiyono.

The two-year studies included flooding of fields soon after applying fertilizer. Researchers used a specialized unmanned aerial vehicle and camera with multispectral sensors to detect corn growth anomaly during the growing season that can lead to yield loss.

The flooded fields caused no loss of yield for corn at the Doyle Chambers Central Research Station in Baton Rouge, while diminishing corn yield at Red River Research Station in Alexandria. However, the results were exactly opposite for flooded soybean fields at each location.

Combined Red Edge and near-infrared spectroscopy technologies were used to scan the field and detect any deficiencies early enough where additional fertilizer could be added to correct potential loss.

“We scan the plants, take the images and apply machine learning algorithms to determine seed counts and whether you have a good or bad stand,” said Setiyono. “We can use this kind of technology to anticipate whether there could be a problem down the road.”

Normalized difference red edge (NDRE) index maps helped the researchers rapidly detect N levels in the fields. Further research needs to determine if rescue nitrogen applications, following the flooding of fields, has yield benefits during variable climate conditions.

One of Price’s projects compared drone nozzle types and how well they infiltrate droplets into soybean canopies. Tests were between regular hydraulic orifice nozzles (similar to boom-type sprayers) compared to newer, more widely used centrifugal rotary spinning nozzles.

“We’ve been finding the spinner nozzles are quite a bit more efficient,” said Price. “They’ll actually get droplets all the way down to the bottom, so we’ve been really impressed.”

Future nozzle tests will compare insecticide coverage and effectiveness on crops.

Another set of spraying tests compared the effectiveness of various spraying heights. Operators typically spray 4 feet or 10-12 feet above the crops.

“We’re seeing as you get the drones higher, you actually get better infiltration with the droplets versus low flying,” said Price.

Another aspect of Price’s drone research has been to detect early disease and insect damage to corn and soybean using RGB and thermal cameras.

“We’re just collecting data now, but we would like to get to where we can use the drone to actually predict the disease severity rating for these plants,” said Price.

Flooded cornfield

Flooded corn field, Doyle Chambers Central Research Station, Baton Rouge. Photo by Tri Setiyono

Aerial photo of a cornfield. There is a ban parked on a side road on the right.

Monitoring nitrogen status in corn using normalized difference red edge (NDRE) index under excessive rainfall. Photo by Tri Setiyono
9/23/2024 8:58:22 PM
Rate This Article:

Have a question or comment about the information on this page?

Innovate . Educate . Improve Lives

The LSU AgCenter and the LSU College of Agriculture

Top