MINNEAPOLIS/ST. PAUL (03/31/2022) – A new technology approach developed by researchers at the University of Minnesota will allow key stakeholders to identify important crop types earlier in the season than ever before.
Satellite imagery has long been used by agricultural agencies to find out what crops are being grown in the field. This allows stakeholders to forecast grain supplies, assess crop damage due to environmental factors, and coordinate supply chain logistics.
While this information is vital, currently available crop mapping products are unable to provide these statistics at the start of the agricultural season. For example, the Crop Data Layer (CDL), a national crop mapping product of the USDA’s National Agricultural Statistics Service, is often not released until four to six months after the fall harvest. This is due to the lengthy process of collecting information from the ground needed to train the main algorithm to separate crops from satellite imagery.
In a study recently published in Remote Sensing of Environment, researchers from the University of Minnesota explain their development of a new method that would allow stakeholders to know where corn and soybean crops are grown as early as July, with precise similar to the USDA CDL, and without the need for ground surveys.
With the rapid growth in the availability of satellite data and advances in artificial intelligence and cloud computing, the bottleneck of satellite crop type mapping has shifted to a lack of labels. ground truth, which are records of crop types at specific locations. In such cases, scientists have attempted to use outdated labels to identify target year crops.
For example, to map crop types in 2022, scientists would develop a model using tags collected in 2021, 2020 or even earlier to develop a model when a new ground survey is not available. or is not feasible. However, this type of model often fails because changes in soil, weather, and management practices in any given year can alter the appearance of crops in satellite images.
To circumvent the need to collect ground tags, the method developed by this research team generates pseudo-tags (they are called “pseudo” because these tags are not collected in the fields) in any target year on the base of historical crop type maps.
This method mimics the way humans identify objects based on their relative positions (also called topological relationships) in an image and uses a computer vision model to identify corn and soybeans based on their topological relationships in a space. two-dimensional derived from satellite imagery. These generated pseudo-tags have a similar quality to tags collected in the field and can be used for the important task of early-season crop type mapping.
“It’s a revolutionary approach that uses computer vision technology to mimic the way humans identify different things in photos. It’s not only fun, but also powerful, as it saves time and labor to conduct field surveys and allows us to accurately predict crop types as early as July,” said Zhenong Jin, Ph.D., assistant professor in the Department of Bioproducts and Biosystems Engineering at the University of Minnesota.
“We found that stable topological relationships existed for different cultures in different years and in different countries, indicating that our approach has the potential to be extended to a general framework that works for many different scenarios,” said said Chenxi Lin, who holds a Ph.D. candidate and first author of the work advised by Jin.
The study also found:
- The approach could generate pseudo-tags of similar quality to field-collected tags for different crops grown in different years and from different regions.
- In the United States, the accuracy of crop type mapping based on the generated pseudo-tags could approach the USDA’s Cropland Data Layer (CDL) product at least six months earlier.
- In northern France, this method can help significantly reduce the number of floor tags needed to produce accurate crop maps, which can be a challenge due to the number of crops grown in the region.
Additionally, the high-quality maps of early-season crop types generated from the proposed approach are also useful for a variety of other activities.
Comprehensive and timely monitoring of insured cropland is beneficial for insurance companies to better design their products. Additionally, crop acreage and production estimation can help commodity traders better project prices and hedge accordingly.
As the researchers look to the future, they recognize that the implementation of this approach relies on enough historical ground-truth labels, which is not a problem for resource-rich regions like the states. States, but is a limited resource for regions like Africa.
However, implementing the approach in underdeveloped countries like many others in Africa could have deeper implications for the ultimate goal of achieving a food secure world. The team plans to extend the framework presented in this study to these regions by incorporating other advanced deep learning algorithms to reduce the need for historical labels.
About the College of Food, Agriculture and Natural Resource Sciences
The College of Food, Agriculture, and Natural Resource Sciences (CFANS) at the University of Minnesota strives to inspire minds, nourish people, and sustainably improve the natural environment. CFANS has a heritage of innovation, bringing discoveries to life through science and educating the next generation of leaders. Every day, students, professors and researchers use science to meet the great challenges of today’s and tomorrow’s world. CFANS offers an unparalleled breadth of experiential learning opportunities for students and the community, with 12 academic departments, 10 research and outreach centers across the state, Minnesota Landscape Arboretum, Bell Museum of Natural History and dozens of interdisciplinary centers.