Increased Situational Awareness

KBR transformed the production of the U.S. National Land Cover Database (NLCD), a key data source used by scientists, resource managers, and planners to inform critical policy and land management decisions across federal, state, and private sectors. Historically, creating the NLCD data package required 2 to 3 years of intense manual processing, which greatly delayed the delivery of critical geospatial intelligence and impacted analyses and strategic policy decisions. By introducing advanced AI/ML techniques such as automated feature extraction and deep learning-based classification, we reduced this production timeline to a few months and slashed processing costs nearly 90%. The result was a dramatic improvement in efficiency, accuracy, and scalability, enabling USGS to deliver timely, high-quality land cover data and accelerate key decision timelines.

We apply advanced computer vision techniques to deblur images, particularly in the context of synthetic aperture radar (SAR) and other complex imagery whose native resolution does not fully resolve features of interest. Using deep learning architectures such as real-valued neural networks, complex-valued neural networks, and custom implementations, we accurately enhance SAR image quality. Our AI-driven approaches significantly outperform traditional autofocus methods, resulting in higher fidelity imagery, improved feature extraction, and more reliable downstream analyses and decision support.