Orbital Insight has launched multiclass object detection algorithms within its GO platform. Multiclass object detection uses advanced computer vision algorithms to identify and differentiate between classes of ships, aircraft and vehicles within satellite imagery. The state-of-the-art multiclass ship detection algorithms will bring immediate benefits to the defense and intelligence community for security, supply chain and other critical needs.
Orbital Insight’s flagship GO platform is purpose-built for the DoD/Intelligence Community to deliver insights in a cloud-agnostic, secure and readily deployable environment. GO will deploy to government networks this year. GO combines information from the world’s sensors—including satellite, aerial, AIS and IoT devices—to analyze economic, societal and environmental trends at scale. The technology supports critical activity-based intelligence by helping spot trends buried within billions of data points, such as patterns at global ports, airports and buildings. The platform is also an invaluable tool for commercial sectors like financial services and real estate to understand evolving trends.
Each of the new multiclass object detection algorithms will help the military and intelligence community automatically monitor activity at thousands of areas of interest. Aircraft detection can find and separate fighters, bombers, helicopters, commercial jets and more; ship detection can identify war ships, cargo ships, fishing vessels and other fleets; and vehicle detection can differentiate between cars, trucks, railcars and soon to be mobile missile launchers and tanks.
Orbital Insight works with several strategic satellite and data partners to ensure customers have access to the latest and most accurate view of what’s happening on and to the Earth. For multiclass object detection, Orbital Insight leverages Planet‘s SkySat constellation and Airbus‘ Pleiades for clear, high-resolution global satellite imagery. The successful algorithms are the result of training the model across hundreds of thousands of images and overcoming challenges such as clustered objects, weather obstacles and low dataset diversity.
Like all of Orbital Insight’s products, multiclass object detection was developed within an ethics framework that shapes the company’s work and values privacy.
"Understanding the movement of specific ships, aircraft and railcars is critical to our national security," said Kevin O'Brien, CEO at Orbital Insight. "Computer vision can make this a reality at a scale and granular level of analysis that humans can't possibly do on their own. Through these multiclass object detection algorithms, our customers will have a clearer picture of activities that have an immediate benefit to our security posture."
"Coupling Planet's 50cm SkySat data with Orbital Insight's advanced analytics provides users with deeper insight into changing conditions on the ground," said Jim Thomason, Planet's Vice President of Imagery and Analytics Products. "We're eager to see the use cases this layered dataset will unlock as users more quickly identify features and make decisions in areas that matter to them most."