Fast Data Processing for Hyperspectral Sensors on Small Platforms

Hyperspectral imaging is a very promising technology for nuclear proliferation detection. However, due to size and weight restrictions, small hyperspectral platforms such as satellites and small drones lack the on-board computing resources for accurate, real-time analysis of the enormous flow of data that a continuously operating hyperspectral sensor generates. This severely limits satellite systems, which can collect far more data than what they can telemeter, and hinders the ability of all platforms to adapt their missions on the fly in response to observations. This proposed program addresses the hyperspectral data processing problem through development of new, fast and accurate algorithms that produce data products in real time. The algorithms circumvent the major computational bottlenecks in existing data processing streams, and would be incorporated in lightweight, power-efficient single-board computer systems. In Phase I, the key algorithms, which perform atmospheric retrieval and correction and target detection and identification, will be coded in prototype form to verify the principles of the computational speedups. Testing will be performed with selections from archival hyperspectral data from a variety of sensors. Finally, a conceptual end-to-end Phase II software and hardware system capable of producing real-time data products will be defined.The toolkit of fast algorithms will be immediately useful in current and future hyperspectral systems being built by the Government and by private industry. These include upcoming satellite constellations that will provide global hyperspectral imagery from space, and drone-based hyperspectral systems. Civilian applications include monitoring for short or long term changes in Earth surface properties such crop health, other environmental monitoring, land management, drug interdiction, disaster preparedness and damage assessment. Military and security applications include border and perimeter surveillance, site monitoring, search and rescue, and disaster response.