IARPA wants to automate geospatial imagery analysis
The intelligence community’s research agency is asking for help developing tools and techniques that would help it monitor construction projects and other manmade activities from multiple space-based or air-borne sensors at global, regional and local scales.
The Intelligence Advanced Research Project Activity issued a broad agency announcement on Feb. 7 for a Space-based Machine Automated Recognition Technique (SMART) that would help it fuse information from multiple data streams and detect and characterize events or activities – from urban development to mudslides -- from an increasing mountain of geospatial intelligence. The objective, according to IARPA, is not object recognition, but rather "broad-area search and categorization of the evolution of events or activities … that can be automatically identified in time-series data."
Unlike today's image analysis algorithms that are generally designed for use with a single data source, SMART's use of spectral data from multiple sources will require harmonization of different data streams and the ability to track changes over time.
SMART would use a pixel-based approach to measure and track changes in surface reflectance across the area of interest over time. It would also map dependencies among stove-piped sensor workflows and be able to perform atmospheric and geometric corrections for each data stream.
IARPA said it expects high-performance computing would be required to accommodate not just the rapid ingest rate of new information, but also support modeling, algorithm development and data-intensive computations.
The agency wants to use the capabilities for a wide variety of applications. Besides monitoring heavy construction projects, it could be used to support disaster recovery and humanitarian aid as well as automated assessment of land-use trending for commercial purposes.
Proposals are due April 7.
Connect with the GCN staff on Twitter @GCNtech.