A technological solution could soon aid soybean producers and crop inspectors in determining the quality of soybeans bound for sale.

Kevin Hoffseth, an engineer who specializes in the analysis of the mechanics and deformation of biological materials, is developing ways to use cameras and computer algorithms to better analyze soybeans.

New technologies will never replace human crop inspectors, Hoffseth said, but they can help them become more consistent.

“It makes everyone’s job easier by removing some subjectivity and improving accuracy and precision in the evaluation of visual indicators of soybean quality,” he said. “You can’t replace the inspection process. It’s worked great for a long time, but there’s nothing to say you can’t help the process.”

In conversations with soybean producers, Hoffseth found some felt frustrated with unexpected variations in grades their crops received in the past. The tools Hoffseth is developing could help producers know what to expect when they transport their crops to the warehouse for sale.

“I think it removes a lot of the stress and uncertainty,” Hoffseth said.

Hoffseth first received grant funding for the soybean imaging project in 2020, and over the past two years his team has worked to build a soybean-specific imaging system from scratch. The system comprises two parts: custom software that can analyze the color, shape and texture of soybeans and machinery that photographs the beans.

Creating a software that can view texture as well as the human eye is challenging, Hoffseth said.

“When you see a pattern on a cushion or the roughness of a sawn piece of wood or even spackle on your wall, it comes back to what people can intuitively figure out,” he said. “With your eyes you can see that the surface is not perfectly flat. You can see the undulations.”

Hoffseth has harnessed the same image processing techniques used in applications ranging from medical imaging to satellite imagery to decipher texture.

“Once you have the image in digital form, you can perform all these mathematical operations on it,” he said. “You can calculate a lot of stuff, and essentially what you’re doing is you’re finding the patterns in the data.”

Using color, shape and texture to analyze the quality of soybeans has shown promise in connecting to existing standards for crop inspection.

“Part of what we’re trying to do is link up to the visual measures that they use as part of that process, which is indirectly linked to the transaction that the producers have with the buyers,” he said. “How can we level that playing field? How can we give a repeatable, accurate sort of measure of a lot of these visual indicators?”

While developing this imaging software, Hoffseth and his team have tested several ways of creating high-quality images of soybeans. Because no sufficient imaging solutions exist, they are building new machines that might be patentable.

“Maybe you can do an assembly line — one by one that is really fast — or maybe you could take one big photo of the top of a truck bed?” Hoffseth said. “There are all these different ways to possibly do it, so we’re looking at ways that have some potential in really speeding up the process.”

Throughout the research process, Hoffseth has spoken with soybean producers to understand their needs, and underlying challenges of the work.

“I want to really address their needs and the questions they have had,” he said. “They have helped me figure out what their needs are, and that helps me think about the possibilities.” Kyle Peveto