The National Geospatial Intelligence Agency (NGA) has announced a significant industry call for assistance in the development of AI-driven computer vision systems, with a potential value of up to $708 million over a span of seven years. This initiative focuses on enhancing the processing of satellite imagery and identifying key targets.
The Sequoia program, established under an indefinite delivery/indefinite quantity (ID/IQ) contract, will enable selected vendors to provide crucial data labeling services. These services are essential for training artificial intelligence and machine learning systems to differentiate between various objects, serving as a foundational capability for NGA’s expansive Maven program.
NGA assumed control of the Maven program from the Defense Department in 2022. The agency is responsible for collecting imagery from satellites and aircraft, analyzing it, and distributing the resulting geospatial intelligence (GEOINT) products, including three-dimensional maps, to various users across the U.S. government, including Department of Defense (DoD) officials and military leaders.
According to the NGA announcement, the ID/IQ contract will support the NGA Maven Program by applying AI capabilities to intelligence, surveillance, and reconnaissance (ISR) sensors and platforms, primarily employing computer vision techniques. The NGA’s computer vision algorithms utilize robust data labeling to execute various automated tasks, including object detection, tracking, classification, and pattern detection.
In addition to computer vision, the Sequoia contract will also facilitate natural language processing, analytical models, and AI/machine learning models that contribute to business process automation within the GEOINT mission.
Vice Adm. Frank Whitworth, NGA director, highlighted the significance of this initiative, stating it represents „the largest data labeling request for proposal in the U.S. government” and indicates a substantial investment in computer vision, machine learning, and AI technologies.
The growing influx of GEOINT data from numerous new satellites and sensor platforms presents a notable challenge for the Intelligence Community. While AI systems are poised to expedite analysis, it is imperative that these models be effectively trained to accurately identify military targets and detect anomalous activities.
