How to distinguish maize from soybean on satellite images
Scientists of the University of Illinois have developed a new technique for distinguishing maize and soybean fields on satellite images using satellite data and the processing power of supercomputers.
The technique was developed to better predict maize and soybean yields for Illinois by the College of Agricultural, Consumer and Environmental Sciences (ACES) from the University of Illinois (USA). It is considered to be a breakthrough because, previously, national maize and soybean acreages were only made available to the public 4 to 6 months after harvest by the USDA (United States Department of Agriculture). The lag meant policy decisions were based on stale data. But the new technique can distinguish the 2 major crops with 95% accuracy by the end of July for each field – just 2 or 3 months after planting and well before harvest.
By using the short-wave infrared (SWIR) band, researchers are now able to distinguish maize and soybean crops on satellite imagery well before harvest. Photo: Kaiyu Guan, University of Illinois
Short-wave infrared band
Yaping Cai, a Ph.D. student working on the project says: “We found a spectral band, the short-wave infrared (SWIR), that was extremely useful in identifying the difference between maize and soybean. It turns out maize and soybean have predictably different leaf water status by July most years. We used SWIR data and other spectral data from 3 Landsat satellites over a 15-year period, and consistently picked up this leaf water status signal.
Kaiyu Guan, assistant professor in the Department of Natural Resources and Environmental Sciences at the University of Illinois and Blue Waters professor at the National Centre for Supercomputing Applications (NCSA) and the principal investigator of the new technology adds: “The SWIR band is more sensitive to water content inside the leaf. That signal can’t be captured by traditional RGB (visible) light or near-infrared bands, so the SWIR is extremely useful to differentiate maize and soybean.”
Deep learning neural network
The researchers used a type of machine-learning, known as a deep neural network, or deep learning to analyse the data and focussed their analysis within Champaign County, Illinois, as a proof-of-concept. Even though it was a relatively small area, analysing 15 years of satellite data at a 30-meter resolution still required a supercomputer to process 10’s of terabytes of data. The researchers are now working on expanding the study area to the entire American Corn Belt, and investigating further applications of the data, including yield and other quality estimates.
To comment, register here
Or register to be able to comment.