Abstract
Abstract. Among the many open-source software solutions recently developed for the extraction of point clouds from a set of un-oriented images, the photogrammetric tools Apero and MicMac (IGN, Institut Géographique National) aim to distinguish themselves by focusing on the accuracy and the metric content of the final result. This paper firstly aims at assessing the accuracy of the simplified and automated calibration procedure offered by the IGN tools. Results obtained with this procedure were compared with those achieved with a test-range calibration approach using a pre-surveyed laboratory test-field. Both direct and a-posteriori validation tests turned out successfully showing the stability and the metric accuracy of the process, even when low textured or reflective surfaces are present in the 3D scene. Afterwards, the possibility of achieving accurate 3D models from the subsequently extracted dense point clouds is also evaluated. Three different types of sculptural elements were chosen as test-objects and "ground-truth" data were acquired with triangulation laser scanners. 3D models derived from point clouds oriented with a simplified relative procedure show a suitable metric accuracy: all comparisons delivered a standard deviation of millimeter-level. The use of Ground Control Points in the orientation phase did not improve significantly the accuracy of the final 3D model, when a small figure-like corbel was used as test-object.
Highlights
The three-dimensional (3D) model of an object, even if it is a sculptural element rather than a more or less extensive portion of territory, is the result of a complex process, starting from data acquisition up to the management and sharing of a PCinteractive 3D virtual reconstruction of the scene. 3D modelling for Cultural Heritage (CH) documentation (Godin et al, 2002) has recently gained large diffusion and its applications range from digital conservation, accurate archiving and 3D catalogues up to cross-comparisons, computer-aided restoration and rapid prototyping
This stereo pair was oriented with 7 well distributed targets of the test-field, chosen as Ground Control Points (GCPs); the inner orientation was performed using the most significant sets of calibration parameters computed with the two procedures
The simplified calibration procedure proved to be satisfactory, further tests will be performed in the future to verify if the combined use of Apero and external GCPs may enhance the accuracy of the calibration final results
Summary
The three-dimensional (3D) model of an object, even if it is a sculptural element rather than a more or less extensive portion of territory, is the result of a complex process, starting from data acquisition up to the management and sharing of a PCinteractive 3D virtual reconstruction of the scene. 3D modelling for Cultural Heritage (CH) documentation (Godin et al, 2002) has recently gained large diffusion and its applications range from digital conservation, accurate archiving and 3D catalogues up to cross-comparisons, computer-aided restoration and rapid prototyping. Active sensors and range data (Blais, 2004; Vosselman and Maas, 2010) offer high level of automation and high geometric accuracy These advantages have made TLS quite common in the last years in the CH field, despite their drawbacks, e.g. high cost, weight, considerable data collection and processing time and a lack of good texture. The research has led to the development of many open-source software solutions for the extraction of point clouds from a set of un-oriented images, e.g. Bundler-PMVS (Snavely; Furukawa and Ponce), Photosynth (Microsoft Corporation), ARC3D (ARC3D), etc In this context, the French mapping agency (IGN - Institut Géographique National) delivered in 2007 several open-source photogrammetric tools with the aim of evidencing their accuracy and the quantitative content of their final results. Apero and MicMac (Apero, MicMac) are the two main tools being distributed: the first is a software for computing orientation of images (Pierrot-Deseilligny and Clery, 2011); the
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
