RC-Heli and Structure & Motion Techniques for the 3-D Reconstruction of a Milan Dome Spire

The paper describes a complete workflow for the 3-D photogrammetric surveying and modeling of complex architectonical objects. The application is focused to the major spire of the Dome of Milan, that is characterized by the well known striking vertical geometry and a high level of complexity. Because no positions from which capturing images and scans exists, a 5 kg payload model helicopter has been used to lift and carry a calibrated digital camera, which has been adopted to capture images for the reconstruction of highest part of the spire. The huge quantity of image data has been partially oriented with a new algorithm able to detect automatically corresponding tie points among the overlapping images by using a Structure & Motion approach, but extended to blocks with a generic shape. The algorithm can identify homologous points with the SIFT operator alternated to a robust estimation of camera poses to remove wrong correspondences. Then a progressive resection alternated with triangulation is carried out to determine the orientation parameters of each image. Finally a photogrammetric bundle adjustment is computed to derive final exterior orientation parameters.