(Vienna University of Technology, Christian Doppler Laboratory for “Spatial Data from Laser Scanning and Remote Sensing”, Austria)

Many applications in cultural heritage management, monitoring and documentation benefit from precise and detailed three-dimensional models. Typical examples would include hazard assessment, planning and reconstruction processes, or public relations. In the past decade, laser scanning has proved to be an adequate instrument for the acquisition of the geometry of cultural heritage. Point clouds acquired by terrestrial laser scanners can be used for the generation of three-dimensional models.
The most accurate models can be determined from data acquired by so-called triangulation scanners. Such scanners typically allow for accuracy with deviations of less than one millimeter. However, their restriction in the maximum measurement distance and their limited field of view is, in general, economically prohibitive, especially for large-scale applications. By contrast, so-called mid-range scanners, based on the phase-shift distance measurement principle, have an almost unlimited field of view and allow for maximum measurement distances of about one hundred meters. However, these scanners are less accurate than close-range scanners.
For the purpose of the geometric documentation of an object before and after restoration, millimeter accuracy is required. Hence, we investigate the magnitude of systematic errors of such scanners based on a practical example. The reasons for such errors are typically deficiencies in the scanning configuration, insufficient scanner-calibration and registration artifacts.
We analyzed point cloud data acquired from a fountain situated in the cour d’honneur of the UNESCO World Heritage Site Schönbrunn Palace before and after a rigorous restoration process. Approximately twenty million points were acquired at each campaign. First, we applied a three-dimensional filtering process to the individual scans in order to minimize the influence of random errors and to reduce the number of points. After the registration of the scans, we analyzed the remaining systematic errors by comparing pairs of points of overlapping scans. Finally, the scans were merged and a three-dimensional triangulation model was generated.
Our experiment has shown that the remaining systematic errors have an extent of approximately two millimeters. Therefore, global differences of the two models can be detected if they exceed two millimeters. For further investigation of these differences, we used color-coded difference models. These models were generated for the whole fountain and additionally for particular parts of it as huge elements were located in new positions.

Keywords: Laser Scanner, Accuracy, Modeling, Cultural Heritage