Robust Hyperspectral Target Reacquisition Under Varying Illumination Conditions and Viewing Geometry

ABSTRACT: Toyon Research Corporation proposes research and development of algorithms which enable significant improvements in target reacquisition based on hyperspectral measurements collected under different illumination and geometric conditions. The proposed work considers challenges due to unknown, non-Lambertian target bi-directional reflectance distribution functions (BDRFs), as well as methods for incorporating partial knowledge of BDRFs, when available. Target handoff between sensors with different hyperspectral sensor characteristics is also considered. To address these challenges, uncertainty modeling and resolution is proposed in Bayesian estimation framework, and hyperspectral image perspective diversity is exploited, as well. The proposed Phase I effort also includes data collection, implementation of prototype algorithms, performance evaluation of the algorithms, and documentation of the results of the research and development. BENEFIT: The successful completion of Phases I and II of this effort will result in development of real-time software which performs target reacquisition using hyperspectral images, with robust performance in the presence of illumination, geometry, and sensor characteristic differences. The effort will also result in documented advances in the understanding of target BDRF characteristics in the hyperspectral domain, as well as models useful for target library development and target classification algorithm development. These technical advances will enable identification of changes in the locations of high-value military targets, as well as persistent tracking of targets in asymmetric and conventional warfare applications. Additional applications include security surveillance in high-traffic-density environments, identification of production or transport of restricted chemicals, and improved analysis capabilities for land use monitoring and environmental protection applications.