Sascha RASZTOVITS
(TU Vienna, Institute of Photogrammetry and Remote Sensing, Vienna, Austria)

Abstract: Two methods are commonly used for the automated generation of 3D documentation models from archaeological assets: Terrestrial laserscanning (TLS) and image matching. Although the usability of processing software for laserscanning data improved significantly in the recent years, this technology still requires a professional operator to be used efficiently. By contrast, for processing digital images taken by consumer (i.e. non-calibrated) cameras, a series of open source software tools as well as freely available online web-services are available, enabling the generation of 3D models from images for non-expert users as well. Those tools perform both required steps, i.e. image orientation generally based on a sparse-matching, and object point determination by means of dense matching. Finally, textured surface models (e.g. triangulations) are generated automatically.
The available photogrammetric tools have benefits and disadvantages due to the ability of coping with outliers, varieties in the dense matching methods applied, occlusion handling etc. Within our study, we evaluate the quantity (i.e. the economic efficiency) and quality (i.e. accuracy and reliability) capabilities of a series of such tools. Different objects and different acquisition configurations are examined aiming at optimizing the configuration for different objects (e.g. façade, sculptures, etc.) with respect to the capabilities of the different tools.
As, in general, the accuracy of 3D models determined from TLS data is better than that of matching results, the accuracy analysis is realized by comparing the matching models to the TLS reference model. Finally, various results (i.e. textured 3D models) of archaeological and cultural heritage sites are shown.

Keywords: archaeology, 3d models, economic efficiency, texture mapping, open source