ARETE ASSOCIATES — Department of Energy SBIR Phase II: 03a

Description

Hyperspectral Imaging is a modality that enables remote detection, identification, and quantification of materials for a wide variety of Government and commercial applications. With modern advances in Hyperspectral Imaging sensors, the number of hyperspectral datasets available for remote material detection and quantification will continue to increase. This program has developed a source-agnostic data analysis framework, “GeoKit,” which has been extended to include hyperspectral data analysis. This will help shorten the to information extraction and delivery, through the development and sharing of workflows. Examples of information that can be extracted from hyperspectral data include agricultural (e.g., crop health analysis), geological (e.g., volcanic gas emission), and non-proliferation activities (e.g., treaty verification). The overall goal of this work is to enable better material quantification from a variety of hyperspectral systems, through advanced algorithms and studies of noise and false alarms statistics to provide robust measurements with confidence intervals associated. Phase I was able to incorporate novelhyperspectral exploitationalgorithmsinto the GeoKit framework. This provides the ability to easily share developedalgorithms, apply themto a variety ofdatasets, or perform large-scale parameter studies. This helped to develop advanced gas detection and quantification algorithms, and provide statistical measures ofperformancewithassociateduncertainties.These algorithms were developed under physics-first principles, and designed to work in the visible through short-wave infrared regimes. They have been successfully applied to long-wave infrared data as well. In Phase II, this program will improve the detection and quantification of gas plumes and incorporate solid target detection. This will require incorporating atmospheric compensation into the GeoKit framework, and studying noise sources in hyperspectral data to better mitigate false alarms. These techniques will be applied to data ranging from visible to long wave infrared, and validated through both simulation and testing on real-world data.