Surface identification from multiband LADAR reflectance with varied incidence angle via database mapping.

Abstract

Incident angle dependencies of LADAR reflection depend on bulk material reflectivity and surface texture properties that can be exploited for surface identification. In this paper, surface identification via multiband LADAR reflected radiance is assessed using the nonconventional exploitation factors data system database. A statistics-based dimension reduction algorithm, stochastic neighborhood embedding (t-SNE), is used to separate the data clouds resulting from the monostatic LADAR reflected radiance and corresponding band ratios. The application of t-SNE to multiband reflected radiance effectively separates the data clouds, making surface identification via multiband LADAR reflectance possible in the presence of unknown incident angle dependencies and uncertainties. It is demonstrated that, for both the multiband monostatic reflected radiance and band ratios, the application of t-SNE mapping yields a significant improvement in surface identification from measurements with unknown or varied incident angles.