D. Watson
Texas A&M University, Texas, United States
Keywords: surveillance, AI, nuclear, satellite
Large-scale surveillance missions require substantial resources and are susceptible to human error. To mitigate these challenges, autonomous methods—specifically anomaly detection and classification using artificial intelligence (AI)—have been employed to enhance human capabilities, enabling planet-scale surveillance. Under the Consortium for Enabling Technologies and Innovation (ETI), funded by the National Nuclear Security Administration (NNSA, NA20), a multi-modal remote surveillance platform utilizing satellites and AI for terrestrial anomaly detection has been developed. This platform focuses on predictive, on-demand characterization of localized anomalies on Earth's surface, subsurface, and atmosphere. By leveraging satellites equipped with advanced sensors/hardware and AI algorithms, the system enhances remote detection of nuclear activities via secondary optical signatures. The signature-based approach provides foundational architectures for facility monitoring. Development efforts include generating specifications for satelite architectures and sensors informed by current technologies, creating surrogate datasets, and compiling a phenomena database for future surveillance applications. Initial results demonstrate successful anomaly detection using low-resolution satellite imagery with convolutional neural networks. Future work aims to mature characterization methods and training libraries using real and high-fidelity surrogate datasets, expand phenomena characterization metrics for efficient orbital data processing, and produce a Front-End Engineering Design (FEED) report for the proposed satellite constellation.