Abstract. In this paper, we present our approach for dense 3D cloud reconstruction using two hemispheric sky imagers with fisheye lenses in a stereo setup. Fisheye lenses follow a different projection function than classical pinhole-type cameras, which provide a large field of view with a single image, but also renders the computation of dense 3D information more complicated, such that we cannot rely entirely on standard implementations for dense 3D stereo reconstruction. In this work, we examine the epipolar rectification model, which allows the use of dense matching algorithms designed for classical perspective cameras to search for disparity information at every pixel. Together with an appropriate camera calibration, which includes internal camera geometry and global position and orientation of the stereo camera pair, we can use the disparity information for dense 3D stereo reconstruction of the a cloud and thus estimate its shape. From the obtained 3D shapes, cloud dynamics, size, motion, type and spacing can be derived and used e.g. for radiation closure under cloudy conditions. We implemented and evaluated the proposed approach using real world data and present two case studies. In the first case, we validate the quality and accuracy of the method by comparing the stereo reconstruction of a stratocumulus layer with the reflectivity observations measured by a cloud radar and the cloud base height estimated from a Lidar-ceilometer. The second case analyzes a rapid cumulus convection in the presence of strong wind shear.
High-performance and tunable stereo reconstruction