Automation in Multi-Image Spherical Photogrammetry for 3D Architectural Reconstructions

Abstract

The derivation of 3D metric information from spherical images for interactive exploration and realistic 3D modeling is receiving great attention due to their high-resolution content, large field-of-view, lowcost, easiness, rapidity and completeness. We present a methodology for accurate 3D reconstruction from spherical (panoramic) images acquired by mosaicking separated frame images captured with a rotating head and a consumer grade or SLR digital camera. In particular we focus the attention on the orientation of the panoramas which is achieved by extracting the necessary tie points with a new fully automated procedure, based on feature matching and robust estimators. Results of the automated panorama orientations and 3D reconstructions of architectural scenes are presented and discussed.