Following the system integration and technical validations of all developed technologies, the SCAN4RECO project has progressed to conducting pilot trials with end users at the end of July 2018.

The functionalities of the Scan4Reco platform were first presented to end-users. These included multi-modal scanning, 3D modelling of the CH object’s structure and appearance, interactive visualization, spatio-temporal simulation, degradation annotation and proposed conservation strategies.

Then users attended presentations of the Scan4Reco system setup. They were able to observe the setup of various sensors, the positioning system of the XYZ stage and the rotational motor as well as a demonstration of their role on the Scan4Reco system.

RFSAT has performed 3D modelling test using two types of 3D modelling engines, PIX4D (offline) and Autodesk (online). Due to lack of proper screening of test objects during scanning from ambient objects, people moving behind and varying light conditions, RFSAT has performed a manual masking (solid white) of the surrounding environment, thus producing two versions of all models, one with and one without masking. As experiments have shown, masking images have improved the accuracy of the produced 3D models. The results of modelling are reported below in chronological order of tests performed. The result of scanning most representative aobjects are presented below.

Object #1: Saint Michael (icon)

The icon of Saint Michael has been scanned using correctly configured application of OF-ADC with Raman scanner disconnected, thus speeding up the overall capture of 70 images to 2 seconds per image taken, in 10 rows and 7 columns. However, images were positioned correctly (as seen on the left in Figure 2), resulting in very high-quality and very accurate models, able to model the geometrical defects in the object at sub-millimetre scales.

3D scan of Saint Michael's icon, camera locations (left), point cloud (middle), 3D model (right)

Object #1: Saint Dimitrios (icon)

The icon of Saint Dimitrios has been scanned using correctly configured application of OF-ADC though due to having a Raman scanner connected, the overall capture of 77 images took over 15 seconds per image in 11 rows and 7 columns. However, images were positioned correctly (as seen on the left in Figure 2), resulting in very high-quality and very accurate models, able to model the geometrical defects in eth object at millimetre scales.

3D scan of Saint Dimitrios's icon, camera locations (left), point cloud (middle), 3D model (right)

Object #1: High Relief (bronze sculpture)

A High Relief is originally a bronze object with ageing patina, which has been 3D scanned using Agisoft Photoscan software with Cultlab system from Fraunhofer-IGD, shown in (a) in the figure below. A 3D printed copy made by Fraunhofer-IGD has been hand-painted by RFSAT artists to reproduce a patina and then manually scanned using a number of modelling engines and scanning methods for comparative analysis, one made with Autodesk engine is shown in (b) below.

Models produced from a 3D printed copy show some overshoots. The one above the top of the object is a result of the overprinting, while at the bottom been caused caused by scanning only a front side of an object. Manual image capture of objects as small as 40 centimetres in length induced additionally significant positioning effects and hence geometrical matching of images had to be performed, causing some inaccuracies, resulting in overshoots and overall reduction of model quality.

3D scan of High Relief sculpture, original object 3D scanned using Cutlab from Fraunhofer-IGD (a) and a 3D scan by RFSAT using Autodesk engine of the 3D printed copy from Fraunhofer-IGD (b)

	Demo video of Automated High-Accuracy Modelling using 3D Photogrammetry
	RFSAT's Gallery3D Application