Development of cloud, snow, and shadow masking algorithms for VEGETATION imagery

Cloud detection is an essential part of the preprocessing chain for various products of the VEGETATION sensor aboard the SPOT satellite. State of the art techniques have been developed to construct a 3 level cloud mask and a binary snow mask. A genetic algorithm has been developed to optimize thresholds on the VEGETATION spectral bands. The increase in performance is dramatic, and improves the quality of VEGETATION products (atmospheric correction, synthesis...). Clouds also cast shadows on the Earth's surface. This can lead up to a 40% bias of the true reflectance of the underlying terrain element. A new and fully automated technique has been developed to provide a cloud shadow mask, based on geometry and radiometry.