Soybean scanning and monitoring setup in Kevin Hoffseth’s lab. Photo by Randy LaBauve

LSU AgCenter researchers are fine-tuning a soybean quality assessment machine that uses shortwave infrared reflectance (SWIR) imaging to capture hidden information about the makeup of harvested soybeans.

For several years, LSU AgCenter engineer Kevin Hoffseth and his team of students have been working on an advanced computer vision-based approach to assist with grading bean quality. This complex imaging system is being designed mainly to determine color, texture, shape and any notable damage to the beans as well as oil and protein content

“We are trying to use this imaging to see what compounds are there,” Hoffseth said. “In addition to what we see on the surface, we want to see deeper into the bulk of the soybean.”

When harvested soybeans are brought to the silo for sale, they are given one of five different grades after inspection and evaluation, primarily based on visually assessed damage categories. Currently, for information on harvested soybean oil and protein content, soybean samples must be sent away for laboratory-based testing. These new AgCenter technologies offer tools to increase speed and reduce variance in evaluating damage in soybeans with comparison to oil and protein content from soybean to soybean.

There is minimal prior application of SWIR area scan imaging for the purpose of soybean evaluation, particularly at the oil and compound level. The SWIR approach is being integrated with existing AgCenter machinery to help standardize and automate the assessment process and reduce load on inspectors and technicians.

“We’ve developed these image processing algorithms for visible spectrum light, but we’re figuring out how to apply these same algorithms to the infrared light,” Hoffseth said.

“It’s practical engineering development that we’re doing this year,” he said. “How can we get the same thing as spectroscopy? How do we go from conventional 2D images to where you get more
spatial information?”

A crucial part of the project’s future success will be the creation of a large database of soybean images used alongside big data applications.

“That becomes super valuable because there’s this immediate need of building a database of our locally grown varieties in Louisiana and understanding the trends,” said Hoffseth. “When people in the industry have these stored images, they can run analysis on them, keep them for future reference and then run them on future applications.”

Another research application of this new technology is to analyze the impacts extreme weather has on the quality of soybeans. Hoffseth is working with a team of other scientists on a project that seeks to better understand the overall damage caused by these disruptive events.

“We’re going to use our approach to determine what sort of grain is produced by plants under extreme weather,” Hoffseth said. “So, as we continue to develop our technology, we can apply it to help out in other research areas like this.”

Louisiana harvested 1 million acres of soybeans in 2024. Although extreme weather events like drought in 2023 and Hurricane Francine in 2024 caused a decline in yields, the biggest financial hit to producers has been the reduced quality of their soybeans.

A patent for Hoffseth’s soybean assessment technologies is now pending. He and his team hope the optimization of this promising technology will ultimately benefit the entire soybean industry.