Researchers at The Ohio State University have devised a revolutionary approach for determining crop health using a combination of drones and machine learning techniques, and utilised it to create a new tool that could help future farmers.
The study, which was published in the journal Computers and Electronics in Agriculture, looks into employing neural networks to help quantify crop defoliation, or the loss of a plant’s leaves on a large scale. Disease, stress, grazing animals, and, more commonly, insect and other pest infestations can all contribute to this devastation.
If left unchecked, entire crop fields can be ruined, decreasing an entire region’s agricultural yield significantly. To tackle this, scientists looked at soybeans, a cash crop that is considered one of the four pillars of world agriculture. Zichen Zhang, the study’s primary author and a graduate student in computer science and engineering at Ohio State, used an Unmanned Aerial Vehicle (UAV), or drone, to acquire aerial photos of five soybean fields in Ohio between August and September 2020. The researchers eventually collected over 97,000 shots that they could classify either healthy or defoliated after cutting each UAV image into smaller images.
“Soybeans are one of the most important agricultural products in the United States, whether it be in exports, or in further food products,” he said. Because soybeans are a major source of oil, food, and protein in many parts of the world, a reduction in soybean production in the United States might have far-reaching implications. However, Zhang’s research, which is one of the first to use non-invasive methods to define large-scale crop health, can aid in determining the likelihood of a production decline due to defoliation.
After manually sorting through the photos, researchers discovered that over 67,000 could be classified as healthy, while nearly 30,000 showed symptoms of defoliation, a ratio of more than 2-to-1. The researchers then utilised this data set to examine the capabilities of various learning algorithms to correctly detect which crops were defoliated and avoid generating false conclusions about healthy soybean crops.
However, after finding that none of the existing learning classifiers could provide the level of precision they desired, the researchers decided to build their own deep learning tool from the ground up. Defonet is the end product, a neural network capable of investigating and correctly answering the study’s initial defoliation questions. “This new architecture is tailored toward this workload,” Zhang said. “It has better performance than currently available tools in accuracy, precision and efficacy.”
Click here to view their paper.
