OneSoil, a Swiss agtech company that builds AI tools for precision agriculture, has released a new version of Global Analytics, its system for mapping fields and identifying crops from satellite images. It is available in more than 45 countries, across Europe and in the US, Canada, Argentina, Brazil, South Africa, Australia and Kazakhstan, according to the company's announcement published on PrecisionAg's Global Ag Tech Initiative.
The new version uses satellite imagery at 2.5-metre resolution to draw sharper field boundaries, so each field is easier to tell from its neighbours. Crop detection has grown from 12 crops to as many as 28, from wheat, maize and soybeans to olives, grapes and coffee. In some markets the models also separate growing seasons; in the US, for example, they identify winter and summer crops.
To recognise the extra crops, OneSoil trained its own machine-learning models on a dataset of about 2 million fields built up over years of work with farmers and agribusinesses. Crop records go back to 2016, so users can follow rotations over time, check what farmers have reported and spot regional trends. The company says identification is typically 90 to 95% accurate in validated markets, depending on crop, region and season.
OneSoil expects different users: agtech platforms that want field boundaries and crop data without building their own detection models, food companies watching their sourcing regions, input makers planning markets, and banks and insurers checking farms when they assess risk.
"We are very proud of the system because it gives businesses a broader view of crop distribution and cultivated areas," said Stepan Zulynskyi, OneSoil's chief executive.
Source: PrecisionAg
Comments
(0)