Novel HSI Sensor

The Remote Detection Program within the Office of Defense Nuclear Nonproliferation Research and Development (DNN R&D) is tasked with developing new technical capabilities to improve the detection, identification, and characterization of special nuclear and radiological materials. In addition to the detection of new threats, the office must also meet nuclear test treaty monitoring obligations. Due to limited access to acknowledged foreign sites, remote detection of illicit behavior is becoming increasingly critical for mission success. Hyperspectral imaging is a modality that enables remote detection, identification, and quantification of chemical activities. By measuring light return from a material, the interaction can be used to decode the material's "chemical fingerprint." Algorithms to separate the source, background, atmosphere, and target spectra allows users to deduce materials and processes used in manufacturing, and even to infer quantities of licit or illicit materials being produced. To address the need for remote chemical detection, identification, and quantification, Areté Associates proposes the development of a hyperspectral image analysis framework. The framework will allow new algorithmic workflows while providing the means for automated processing of large datasets. Phase-I will utilize Areté’s atmospheric simulation and gas detection capabilities from previous work to develop two novel new HSI algorithms: a gas quantification algorithm, and “co- indicator analysis.” The algorithms utilize decision theory, statistical analysis, and a deep phenomenological understanding of spectral sensors and material properties to ensure they are robust to noise and output confidence metrics. The novel algorithms will enable DNN’s mission by automatically extracting intelligence about material production, manufacturing processes, and production rates. These are all enabling intelligence for remote treaty monitoring, especially in sites with restricted and limited access. The analysis framework development will be accelerated by the existence of a computational framework for processing, GeoKit, currently under development by Areté under an active Phase-II SBIR. GeoKit has been developed to take multi-source data and provide a fusion capability, by abstracting the data source from remaining algorithm workflow. This decreases the timeframe from new algorithm conception to active processing on large datasets. This increase in efficiency allows for agility and responsiveness to new data sources, methods, and needs. Phase-II and beyond could utilize GeoKit’s fusion capability to utilize auxiliary information to further improve algorithm confidence and intelligence extracted.