An artificial woodland scene model for simulation of hyperspectral imaging system

Hyperspectral images have been used in many applications including object detection, land cover classification and material identification. By building simulation models of hyperspectral system, the key contributors to system performance might be found. Scene model is the most complex subsystem in the hyperspectral simulation system. However, through various simplifying assumptions, simulation model of hyperspectral scene could be built. A scene model based on Monte Carlo simulation approach has been presented by Kerekes. In support of some special scene, such as a park with low shrubbery and high building walls around, Kerekes's hyperspectral simulation model is extended to an artificial woodland scene model in this paper. Bidirectional reflectance distribution functions (BRDFs) has been introduced into the new model and the reflectance functions for the ground materials are represented by combination of canopy BRDFs and subpixel fraction model. The new model also introduces secondary illumination based on geometric analyzing method that can be used to simulate the effects of reflected solar radiance from other building. To study the performance of the model, some validation experiments have been implemented and reasonable results have been achieved.