Atmospheric and shadow compensation of hyperspectral imagery using voxelized LiDAR

Hyperspectral images have become more prevalent and accessible in recent years. Many of the processes which take advantage of hyperspectral imagery rely on the generation of a reflectance product in order to compare with spectral signatures, and numerous methods have been developed to perform this process. However, per-pixel illumination effects are typically not accounted for because the necessary geometric parameters are not available in a 2D image. LIDAR can be used to derive the needed geometric terms to compensate the image for illumination differences and shadows, and this paper demonstrates how the use of voxel 3D modeling can make this process more accurate. We apply our methods to data from the publicly available SHARE 2012 dataset, and show how target detection can be improved by using illumination-compensated reflectance.