Agricultural agencies have used crop mapping through satellite imaging to forecast and assess crop damages, grain supplies, and supply chain logistics.
However, the crop data layer (CDL) United States Department of Agriculture Statistics Service uses for mapping requires a ground survey that usually takes four to six months of data collection before crop separation in the satellite image can be made.
Researchers from Minnesota developed a satellite imaging method to enable early forecasting and crop assessment without the tedious field data collection. Instead of collecting data from the field to generate ground labels, the imaging method generates labels from historical crop-type maps. And then, with computer vision technology, crops from the satellite image are identified through topology relations.
The topology position is similar to how humans perceive relative situations between two objects in a photo. Topology computer technology identifies and separates corn and soybeans from satellite images.
The study suggested that the crop-type map generated by the computer-vision model has a similar quality to the maps with field-collected data. The new method saves time and labor and also accurately predicts the crop early.
Another approach for using this technology is the monitoring and planning of crops by insurance companies. Aside from insurance companies, traders can also use the technology to determine the supply and price trends of the commodities.
This method of satellite imaging has great potential for global use, especially in achieving food security. However, the technology is hampered by the limited crop historical data in other countries like Africa.