From pixel to mesh - accurate and straightforward 3D documentation of cultural heritage from the Cres/Lošinj archipelago

Abstract

Let’s face it: most people like 3D visualizations. Whether it is in movies, holograms or games, 3D (literally) adds an extra dimension to conventional pictures. However, 3D data and their visualizations can also have scientific archaeological benefits: they are crucial in removing relief distortions from photographs, facilitate the interpretation of an object or just support the aspiration to document archaeology as exhaustively as possible. As archaeology is essentially a spatial discipline, the recording of the spatial data component is in most cases of the utmost importance to perform scientific archaeological research. For complex sites and precious artefacts, this can be a difficult, time-consuming and very expensive operation. To date, different institutions and practitioners from a variety of scientific fields focus on the production and usage of 3D models and orthophotos. Many technologies and possible approaches exist, but all are characterised by certain disadvantages. Standard survey methods using a GNSS (global navigation satellite system) receiver or total station are often too time-consuming and generally do not provide the required data density to enable a highly detailed archaeological investigation. Terrestrial laser scanning offers a viable approach to the problem, but the data acquisition is quite expensive while the processing often necessitates a thorough technical understanding and appropriate (i.e. costly) software. In this contribution, it is shown how a straightforward and cost-effective hard- and software combination is used to accurately document and inventory some of the cultural heritage of the Cres/Lošinj archipelago in three dimensions. First of all, standard photographs are acquired from the site or object under study. Secondly, the resulting image collection is processed by means of some recent advances in computer technology and so-called Structure From Motion (SFM) algorithms, which are known for their ability to reconstruct a sparse point cloud of scenes that were imaged by a series of overlapping photographs. When complemented by multi-view stereo matching algorithms, detailed 3D models can be built from such photo collections in a fully automated way. Moreover, the software packages implementing these tools are available for free or at very low-cost. Using a mixture of archaeological case studies, it will be shown that those computer vision applications produce excellent results from archaeological imagery with little effort needed. Besides serving the purpose of a pleasing 3D visualization for virtual display or publications, the 3D output additionally allows to extract accurate metric information about the archaeology under study (from single artefacts to entire landscapes).