A FUSION-BASED WORKFLOW FOR TURNING SLAM POINT CLOUDS AND FISHEYE DATA INTO TEXTURE-ENHANCED 3D MODELS

Abstract. Mobile mapping systems are increasingly developing ad hoc solution and integrated approaches for rapid and accurate 3D digitization in different operating environments belonging to built heritage assets. The use of emerging compact, portable and low-cost solution for imaging and ranging well fits in the purposes of mapping complex indoor spaces especially for narrow and underground ones (tunnels, mines, caves and ancient spaces), that are very challenging contexts in which to experiment integrated technological solutions and tailored workflows. In these cases, the main key issues are generally the difficulty in the seamless positioning and the complete and successful metric-radiometric content association in metric surface, due to the reduced manoeuvring space and complex lighting conditions. The prevalent goals for which the 3D digitization could be conceived are, beyond the accurate metric documentation, the analysis of mutual relations of volumes in complex structures, the virtual reconstruction and navigation of spaces with reduced accessibility for dissemination aims. The new SLAM-based positioning solutions implemented in some recent portable systems for indoor/outdoor mapping are increasingly developing and favoured by geometric features extraction algorithms even in traveling through complex and irregular environments. In parallel, the possibility to exploit the advances in digital photogrammetry algorithms for image matching and dense reconstruction using action-cam, compact and fisheye cameras allows to deploy investigation solutions even in complex environments at first sight impossible to map by photogrammetric approach. Here within the F.I.N.E. benchmark in the site of the San Vigilio Castle (Bergamo) and the “nottole” tunnels, a fusion-based workflow is proposed. It is focused on the purposes of providing radiometrically enriched 3D data from the possibility to colourized ZEB point cloud and a textured mesh surfaces with an oriented image block, taking care of the time processing steps optimization.