Satellites produce 1000's images daily. With increasing resolution and shorter frequency, satellite imagery can now be used to analyze a number of near-real-time changes. Simply classifying each square of land as vegetation, road use, farmland, or water is becoming enormously helpful for climate scientists, planners, and the intelligence community. But how to deal with this overwhelming volume?
Classifying land usage from satellite imagery is revolutionizing the way we approach national security, economic analysis, and environmental impact. With an almost constant stream of satellite imagery becoming available from both commercial and government sources, analyzing that imagery at scale can realistically only be done by computers.
Building on the existing work we’ve done for image analysis at the retail level, Oil & Gas, and healthcare, we are now partnering with Masego Inc to deploy Interplay® and it’s flexible, dynamic, and engine-agnostic AI capabilities for satellite image analysis.
For this project, we used Interplay to rapidly prototype a data flow that could accept a standardized training data set, train the engines, optimize, then output a simple reporting mechanism that shows the image being analysed, resulting categorization, and degree of accuracy (almost always above 99%).
Client:
Military
Goal:
Identify previously undetectable objects
Development Time:
4 weeks
Deployment:
Undisclosed