3d Building Facade Reconstruction under Mesh Form from Multiple Wide Angle Views

In this paper, we present an algorithm, which automatically generates textured meshes of building facades from a set of multi overlapping calibrated and oriented images. We are in a context of massive data production and aim at high geometric accuracy. The central idea is to create a 3D point cloud. After comparison between object-space and image-space techniques, we choose the latter and compute the depths of the pixels in each image using a correlation-based method. A post-processing step is necessary to filter points according to a confidence index and remove the sparse speckle noise. We then perform a global optimization to find a regularized surface. Finally, the 3D point cloud is triangulated. The resulting mesh is an accurate representation of the facade surface from each image view point. We obtain promising results, with a correct texture projection on the reconstructed model.