Material Reflectance Retrieval in Shadow Due to Urban Vegetation from 3D Lidar Data and Hyperspectral Airborne Imagery

摘要

Reflectance retrieval is a key parameter for land cover mapping from hyperspectral imagery. However most of the inverse methods to estimate these reflectances are limited in urban areas with the use of high spatial resolution sensors because they do not take into account the 3D radiative impact of the urban environment. A recent tool, ICARE [1], is able to retrieve surface reflectance in the reflective domain (0.4-2.5μm) in the sunlit and shadow areas overcoming both slope and environmental effects. Its main inputs are atmospheric conditions and the 3D vector model of the scene. This model has proven to perform with good accuracy in shadowed areas cast by opaque structures. Nevertheless ICARE has never been tested in the shadow of vegetation because the 3D information was not available. In this paper, a new dataset including 3D lidar (0.25m) and hyperspectral (CASI 0.5m) data acquired over Norrkoping is processed demonstrating the potential of ICARE to retrieve the surface reflectance over any type of shadows. The results will be assessed by comparing the retrieved reflectance of a given material both in the sunlit and shadow areas. Further, the gain brought by ICARE compared to a flat scene assumption reflectance retrieval method will be evaluated in terms of classification performances.