Improved Hyperspectral Target Detection Utilizing Coincident 3-D Data

This work endeavors to develop a simple model for estimating and correcting the spectra across scene lighting conditions. Much like the atmospheric corrections in MODTRAN, we believe that localized hyperspectral (HSI) lighting correction can be performed independent of the HSI material identification algorithm and can significantly improve performance by reducing the variance of both target and background spectra. Relying on physics, ray tracing, and a simple model for the bi-directional reflectance distribution function (BRDF), we hope to eliminate a reliance upon tunable parameters and algorithmic complexity to produce a simple, yet elegant, algorithm that will generalize to unseen data.