Machine learning helps determine the health of


COLUMBUS, Ohio – Using a combination of drones and machine learning techniques, researchers at Ohio State University recently developed a new method for determining crop health and used it to create a new tool that could help future farmers.

Published in the journal Computers and electronics in agriculture, the study investigates the use of neural networks to help characterize crop defoliation, or the widespread loss of leaves on a plant. This destruction can be caused by disease, stress, grazing animals and, more often, by infestations of insects and other pests.

Left unchecked, entire crop fields can be damaged, drastically reducing the agricultural productivity of an entire region. To combat this, the researchers chose to analyze a cash crop considered one of the four staples of global agriculture: soybeans.

Between August and September 2020, Zichen Zhanglead author of the study and graduate student in computer science and engineering at Ohio State, used an unmanned aerial vehicle (UAV), or drone, to take aerial images of five soybean fields in Ohio. After cropping each UAV image into smaller images, the team ultimately had more than 97,000 photos that they could label healthy or defoliated.

“Soybeans are one of the most important agricultural products in the United States, whether in exports or other food products,” he said. According to USDA, the United States is the world’s largest producer of soybeans and its second largest exporter. Yet domestic farmers are racing to meet demand: Last year, more than 90 million acres of soybean crops needed to be planted to meet consumer demand.

Since soybeans are an important source of oil, food, and protein in many parts of the world, a potential decline in soybean production in the United States could have profound consequences. But Zhang’s study, one of the first to use noninvasive technologies to characterize crop health on a large scale, can help assess the likelihood of reduced production due to defoliation.

“Soybean defoliation is a very common problem, but it’s a problem we can solve,” Zhang said.

After manually sifting through the collected images, the researchers found that about 67,000 of them could be labeled healthy, while nearly 30,000 showed various signs of defoliation, a ratio greater than 2 to 1. they used this dataset to compare the ability of several learning algorithms to correctly infer which crops were defoliated and to avoid making incorrect assumptions about healthy soybean crops.

But after concluding that none of the learning classifiers could deliver the accuracy they wanted to achieve, the researchers decided to build their own deep learning tool from scratch. This final product is called Defonet, a neural network capable of investigating and correctly answering the initial defoliation questions of the study. “This new architecture is suitable for this workload,” Zhang said. “It performs better than currently available tools in terms of accuracy, precision and efficiency.”

If adopted in the field, Defonet could transform agricultural industry decision-making in the event of major crop losses, study co-author says Christopher Stewartassociate professor of computer science and engineering.

“In the years to come, we’re going to have to dramatically increase food production in order to meet demand,” Stewart said. “The idea behind digital farming is to use computing and other technologies to ensure that every seed planted is grown as efficiently as possible.”

The study was also co-authored by Sami Khanal, assistant professor of food, agricultural and biomedical engineering, Amy Raudenbush, research associate of entomology, and Kelley Tilmon, associate professor of entomology. This research was supported by the National Science Foundation.

Contact Person: Zichen Zhang, [email protected]Christopher Stewart, [email protected]

Written by Tatiana Woodall, [email protected]

Disclaimer: AAAS and EurekAlert! are not responsible for the accuracy of press releases posted on EurekAlert! by contributing institutions or for the use of any information through the EurekAlert system.


Comments are closed.