Scientific Objectives

Based on the acknowledged expertise of the technical partners of the consortium, a core set of techniques for 3D matching, retrieval and metrology have been transformed into specialized predictive reconstruction technologies targeting the following objectives:

  • On-the-fly auto-completion for 3D digitization. The shape of digitized CH objects can be potentially predicted during acquisition, based on the gradually available partial scans of an object. The stream of input point cloud data from the acquisition source is used to interactively retrieve and fit the closest matching candidate shape from parts of digitized artifact repository models as well as template models (categorized primitive objects), onto the acquired geometry, thus predicting and automatically suggesting the geometry for the parts not yet scanned. As the acquisition data are the starting point for all subsequent processing steps, their quality and reliability are significant issues. Corrections and fine details can be locally applied, where necessary, using localized complementary scans, effectively minimizing the overall time and cost of the scanning procedure or eliminating the need to post-process the data or attempt scanning in hard to reach surfaces of the original CH artifact. 

    CH objects are especially good candidates for such a system since they can often be categorized, possess regularity, symmetries or repeated patterns and salient features. Furthermore, typical acquisition cases involve immovable, large or heavy to lift parts and fragments, which can be digitized in place, since inaccessible parts could be predicted though auto-completion. 

  • Estimation and prediction of monument degradation. Based on present-time surface shape, material measurements and environmental data, the project investigated highly-efficient techniques for forward and inverse deterioration prediction. This allows to essentially move the artifact's surface condition "forward in time" and visualize the dynamic state of the deteriorating object, in the context of geometric and textural alterations. In order to include  geometric information in the simulation model, PRESIOUS conducted a number of timed, high-accuracy differential surface scans on the degrading monuments. Also, using the digitized data of the monument in its current state, similar surface regions have been retrieved and fitted to the degrading surface. This acts as an additional constraint for the simulation by providing an indication of the intact state of the object.

  • 3D CH fractured object restoration and completion (missing parts synthesis). By exploiting existing CH objects in an example-based object restoration process, automated procedures have been developed for fractured artifact reassembly in three dimensions. This resembles the solution to a three-dimensional puzzle, where the pieces are either intact or broken artifacts retrieved from relevant CH object repositories and the target result is predicted from approximate model templates, which act as constraining guides. The developed techniques rely on shape analysis and processing techniques, similar in nature to the predictive scanning problem. Subsequently, novel techniques for the recovery (prediction) and automatic geometry generation of missing elements were developed; the missing elements prediction is conducted at multiple levels of detail (general shape, detail sculpting) and can thereby aid the physical repair process of the actual objects.

 Scientific objectives and RTD work packages