Radiometrically Accurate Spatial Resolution Enhancement of Spectral Imagery
Abstract
Recent technological advances have seen a proliferation of both airborne and spacebased imaging systems in both the commercial and government sectors. These systems have historically been used primarily for visualization of surface features in either panchromatic (i.e., black and white) imagery or “traditional” color imagery in blue, green, and red. Some systems include a Near Infrared band to perform simple vegetation analysis. Such systems are referred to as multispectral imaging systems. However, many new systems include enough spectral diversity and coverage to allow for quantitative analysis of their collected imagery through automated workflows based on spectral signatures of materials on the observed surface. Such systems include multispectral imaging sensors that include up to eight (8) spectral bands in the Visible –Near Infrared (0.4 um – 1 um) spectral range as well as spectral bands beyond the visible spectrum in the short wave infrared (1 um – 2.5 um). Additionally, hyperspectral imaging systems that incorporate a dispersive element in the optical system image in hundreds of spectral bands, allowing specific material identification. However, to achieve this increased spectral coverage while maintaining high signal-to-noise ratios, the spatial resolution of MSI and HSI systems is generally lower than panchromatic imagers. With the goal of doing quantitative spectral exploitation from multi- and hyperspectral sensors, at ever finer spatial resolutions, there is a need to understand the limits and capabilities of spectral resolution enhancement for these applications, and to develop novel approaches to this problem. In particular, new methods must be developed and investigated that preserve the radiometric fidelity of the spectral sensors that allow analysts to perform quantitative material identification and characterization through sometimes subtle spectral features. The proposed NURI project seeks to understand the capabilities and limitations of current algorithmic approaches and to develop new methods and algorithms for this problem with particular emphasis on understanding their dependence on scene and sensor characteristics. The guiding assumption is that there is not necessarily one algorithmic approach that will be optimal for all scenes and all sensors. The goal is to identify a workflow that will allow us to estimate, a priori, scene characteristics that are highly correlated to optimal approaches among a variety of options, providing a pathway to resolution enhancement with high confidence in the resulting spectra. Additionally, we seek to identify novel approaches leveraging large scale material mapping from commercially available sensors, including temporal changes, to develop new methods for spatial resolution enhancement that ensure spectral fidelity of the processed scenes.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 06, 2020
- Source ID
- HM04761912007
Entities
People
- David W. Messinger
Organizations
- National Geospatial-Intelligence Agency
- Rochester Institute of Technology