Digital Reconstruction of Fragmented Cultural Heritage Assets: The Case Study of the Dacian Embossed Disk from Piatra Roșie

Abstract

The most peculiar characteristic of a cultural heritage is represented by its uniqueness. To ensure that an object is preserved against environmental deterioration, vandal attacks, and accidents, modern Cultural Heritage documentation involves 3D scanning technologies. In the case of fragmented artefacts, the digitization process represents an essential prerequisite for facilitating an accurate 3D reconstruction. The aim of this research paper is to present a framework that enables an accurate digital reconstruction of fragmented or damaged artefacts using ornament stencils obtained from 3D scan data. The proposed framework has been applied for the richly adorned ornaments of the fragmented Dacian embossed disk from Piatra Roșie. The case study makes use of the 3D dataset acquired, using a structured light scanner to extract vector displacement maps, which are then applied to the 3D computer-aided design (CAD) model. The output of the framework includes a proposed digital reconstruction of the aurochs fragmented Dacian embossed disk, as well as the ornaments’ stencils database. The proposed framework addresses problems that are associated with 3D reconstruction processes, such as self-intersections, non-manifold geometry, 3D model topology, and file format interoperability. Finally, the resulting 3D reconstruction has been integrated within virtual reality (VR), augmented reality (AR), and mixed reality (MR) applications, as well as computer-aided manufacturing (CAM) based on additive manufacturing to facilitate the dissemination of the results.

1. Introduction

Modern Cultural Heritage documentation involves interdisciplinary research teams from various fields. The three-dimensional documentation of both cultural heritage sites and artefacts has become a common task in archaeological and restoration processes [1]. This digital documentation process that makes use of modern scanning equipment and software enables the acquisition of precise digital models of both the archaeological sites and artefacts to define digital replicas. Three-dimensional scanning is frequently used for the documentation of Cultural Heritage (CH), as it is capable of defining accurate measurements compared to other digitization methods (photogrammetry, as an example) [2]. The accuracy evaluation of photogrammetry has been analyzed using a custom machine gauge [3] and a coordinate measuring machine (CMM). The trueness and precision of photogrammetry has been analyzed using CMM in other research domains, such as dental implants [4] and industrial applications [5].

Digital Cultural Heritage documentation involves a wide variety of technologies that provide heterogeneous data regarding file structure and file format. Because of this, depending on the final output of the 3D model, there are various standard file type conversion steps that are required to ensure that the output 3D dataset can be integrated within either digital applications such as virtual reality (VR), augmented reality (AR), mixed reality (MR), or computer-aided manufacturing processes (CAM).

One of the most important challenges related to the preservation of cultural and historical heritage is associated with the rapid decay of exterior monuments due to environmental deterioration, as well as vandal attacks [6]. Unfortunately, in many areas around the world, archaeological heritage has been affected by both natural hazards and anthropic destruction. The negative impact of social non-involvement in the protection of archaeological sites has had an impact in multiple countries. Romanian archaeology did not benefit from a specific legislation from its beginning, and until January 2000, as presented by Dorel Micle, even though Romania has been a member of the European Convention concerning archaeological heritage (signed at La Valetta on 16 January 1992) ever since 1997, it was only in 2000 that the government enacted the first law concerning the protection of archaeological heritage. This had an impact regarding the prevention of archaeological treasure hunting, but as presented within this research article aimed at the promotion of the Dacian ornamental shields, various Dacian archaeological sites have been targeted by treasure hunters with metal detectors up until this period, as well as after 2000, in remote areas such as the Dacian Hillfort from Piatra Roșie [7].

Due to various degradation aspects previously presented, CH artefacts are frequently broken into fragments when they are discovered. The traditional manual reconstruction may induce additional damage to the CH fragmented artefacts; therefore, the use of computer-aided techniques for reassembly and digital reconstruction provides several advantages over traditional techniques [8]. The main advantage is represented by the output 3D models that can be used for a wide variety of applications, both digital, such as: VR [9], AR [10], and MR [11], but also regarding the development of tangible replicas using computer-aided manufacturing (CAM) and rapid prototyping (RP) techniques such as casting [12], additive manufacturing [13], laser synthetization, and laser cladding [2].

As presented by Parfenov V. et al., in this context, the need to reconstruct damaged and stolen cultural heritage objects, and to gradually replace them with replicas represents one of the most important aspects on the agenda. This idea of replacing originals with copies is not new, and it causes an ambiguous attitude in society. As an example, the authors have presented the famous masterpiece of Michelangelo—a marble sculpture of David, which has remained for almost 370 years in Piazza della Signoria in Florence, and was copied in 1873 when the original statue was moved for eternal storage in the Gallery of Academy of Fine Arts in the city [6].

Modern technologies that integrate 3D scanning technology and CAM represent an attractive alternative to the traditional manual reconstruction approaches. The main advance of 3D scanning is that it is a contactless method that does not have any negative impact on the original cultural heritage elements [14].

With a wide variety of 3D scanning technologies, determining which solution is optimal is influenced by numerous aspects that are related to the digitization of objects or monuments. As presented by Di Angelo L. et al., there is no 3D scanner on the market that meets all of the technical and economic constraints; therefore, they have developed an Analytic Hierarchy Process method to determine the best 3D scanner for cultural heritage applications [15].

Faithful 3D reconstructions can be defined using 3D scanned datasets of various objects and monuments if there are enough information to define the whole 3D digital reconstruction. To facilitate the dissemination and promotion of digital 3D reconstructions, researchers are also developing VR, AR, and XR environments. The researchers Liritzis I. et al. have proposed the development of an educational application that combines science, history, and archeology to enhance learning regarding the Sanctuary of Delphi [16]. The pilot study was aimed to introduce cultural heritage and archaeology to university syllabuses to support e-learning studies.

As presented by Gomes L. et al., both innovation in data acquisition sensors and an increase in computational power have made digital reconstructions based on 3D scanned models an ongoing research field [17] with many existing challenges regarding current 3D reconstruction frameworks, such as:

• The accuracy and trueness of the proposed digital reconstruction [18];
• 3D modeling problems (self-intersections and non-manifold geometry) [19];
• 3D model topology (triangles or quads) [20];
• File format interoperability towards VR, AR, XR, and CAM [21];
• Aspects related to open and reusable digital cultural heritage 3D models [22];
• Empowering linked Cultural Heritage data [23].

The aim of this research is to address these challenges associated with existing 3D reconstruction frameworks. The accuracy and precision of the 3D reconstructions are resolved using 3D scanning and reverse engineering techniques. The aspects related to 3D modeling problems, 3D model topology, and file format interoperability are addressed by the tools and the proposed ornament extraction workflow. The aspects relating to open and reusable digital 3D models, and linked metadata are addressed by the developed digital applications. The case study presented within the paper has been applied on the well-known fragmented Dacian embossed disk found at the Dacian hillfort of Piatra Roșie, which is currently exhibited at the National Museum of Transylvanian History.

The following section provides a background regarding the finding location of the fragmented embossed disk, and some aspects regarding the iconography, as well as related 3D reconstruction pioneer and state-of-the-art research projects and frameworks. The materials and methods sections present the proposed framework used to recreate the fragmented 3D artefacts, using ornament stencils extracted from the 3D scan dataset, as well as the 3D measurement and modeling of the support model of the disk. The results section presents the 3D reconstructed embossed disk, and the development of the online ornament database, as well as the VR, AR, XR, and 3D printing aspects aimed to facilitate the dissemination and promotion of the Dacian embossed disks.

2. Background and Related Works

2.1. Background Regarding the Dacian Hillfort of Piatra Roșie

The Dacian Fortification from Luncani—Piatra Roșie was built on a rocky hilltop, 817 m above sea level on a peak, on the Luncanilor Valley. According to some historians, the hillfort was built during the reign of King Burebista, and it dominated the Luncani Plateau for a century and a half, until it was destroyed by the Roman legions on their way to Sarmizegetusa Regia. After it was destroyed, its ruins remained unearthed for centuries without anyone ever rebuilding them.

“The first Dacian embossed disk was unearthed in the summer of 1949, by a team of archaeologists led by Constantin Daicoviciu” [24]. During this archeological campaign, discovery was made of several weapons, tools, ceramic pieces, an ancient chandelier, and a bronze mask that would have depicted the goddess Bendis. However, the most remarkable piece were the remains of the mysterious disk. The exact location of the disk was the “in the apse building on Terrace I, in the south-western corner of the room, attached to the corner” [24].

The archaeological campaign of 2021 led by Răzvan Mateescu from the National Museum of Transylvanian History from Cluj-Napoca was performed on Terrace I, where the apse building was partially investigated in 1949 by Constantin Daicoviciu. The two areas, SP I/2021 and SP II/2021, indicated in Figure 1c, have been initiated within the 2021 archaeological campaign. According to the archaeological report, the material found comprises “potsherds, decorated iron tacks and the fragment of a bronze vessel”; all of the items entered the collection of the Museum of Dacian and the Roman Civilization from Deva.

To better highlight the finding location of the fragmented disks from 1949, the authors have defined a 3D model of the Piatra Roșie hillfort based on the topographic map within CATIA V5 software; the resulting 3D model is presented in Figure 1a.

The discovered fragments have been arranged to form an ovoid shape shield with a large diameter of 73 cm and a small diameter of 60 cm, as was noted by the archaeologist Daicoviciu C., in the monograph of the Dacian Fortress from Piatra Roșie [24] published after the archeological campaign. The shield has been conserved ever since within the National Museum of Transylvanian History from Cluj-Napoca, Romania.

As presented within related research articles for half a century since the discovery of the embossed disks fragments, the ruins of the Dacian fortress from Piatra Roșie have not been systematically investigated. However, treasure hunters were interested in its ruins, as well as the ruins of the other Dacian fortresses [25]. In the early 2000s, treasure hunters have found several similar pieces here. As presented by Augustin Lazăr, several disks have been found in the Orăștiei Mountains, and according to the investigators in charge of the Dacian treasure processes, some of them have reached the black market of antiquities. In 2003, a person who was trafficking heritage objects introduced two Dacian embossed disks onto the legal antique market, and in 2004, he sold them to a dealer in New York, USA. The two artefacts were recovered in 2011 by the Romanian authorities; one of them has an aurochs representation and the other one has a griffin as the central ornament [26].

The Dacian embossed disks are some of most precious archeological treasures discovered in the Dacian hillfort of Piatra Roșie in the Orăștiei Mountains, as they are richly adorned with zoomorphic representations. Some embossed disks appeared on the black market of antiques, and the Romanian state made great efforts to recover them in 2011, while other the two Dacian embossed disks are still being pursued internationally, through Interpol.

The current physical reconstruction of the fragmented disk, along with various observations regarding the embossed disk, is presented in detail by the authors Gelu Florea and Liliana Suciu, within an article published [27] in 1994, in Romanian.

The iconography of the fragmented embossed disk is presented in detail within the catalogue entitled “Incursiuni Dacice în Spațiul Virtual”, which is available both in Romanian and in English [28]. As presented within the catalogue, the embossed disks are the true masterpieces of the art smith craft of the Orăștie Mountains. These disks are generally known as “the Piatra Roșie shields”. Crafting them implied advanced technological knowledge and a remarkable degree of artistic talent. The images were embossed by hammering the reverse of a heated iron sheet, and the details were engraved on the obverse side using special instruments. The circular composition is balanced: a central medallion with the most important image, a zoomorphic one (a griffin, a lion, a deer, an aurochs, or a wisent), framed using a secondary decoration with plant patterns (acanthus leaves and/or nested lotus petals). The accuracy of the outlines, the realism of proportions, and the fine details are extraordinary, especially bearing in mind the specificity of the technology employed.

2.2. Related 3D Reconstruction Cultural Heritage Projects and Frameworks

The development and wide adoption of 3D scanning technologies has enabled 3D digitization and reconstruction projects to emerge around the world. In this subsection, we discuss the pioneer projects, as well as the most complex 3D reconstruction projects.

The most prominent pioneer project focusing on the digitization of cultural heritage is the “The Digital Michelangelo Project” by Levoy et al. [29], which integrates the 3D scanning of 10 statues created by Michelangelo. The authors have adopted a wide variety of scanning solutions, including triangulation laser scanners, time-of-flight laser scanners, and digital cameras. As was presented on the project website during their academic year abroad in Italy of a large team including 30 faculty, staff, and students from Stanford University and the University of Washington, the authors have been involved in several side projects such as the Digitizing of the Forma Urbis Romae. The team has scanned the 1163 fragments of the impressive Marble Plan, which is the key source document of ancient Roman topography, measuring 60 feet across, 45 feet high, and 2–4 inches thick.

The authors have developed computer algorithms to aid in the reconstruction of the Marble Plan, and they have created a public database that includes all of the 3D scanned fragments intended to be used as a public tool for study and research. The authors have managed to approximate only 10% of the original monument, based on the features and fitting between fragments. The results have been presented and published in the Proceedings of the Third Williams Symposium on Classical Architecture, Journal of Roman Archaeology, in 2005.

In this case, the accurate 3D reconstruction of the Forma Urbis Romae represents an impossible task, as the existing fragments represent only about 15% of the original marble map, and with the disintegration of the Roman Empire in the fifth century A.D., the Forma Urbis suffered the same fate as the rest of the city; for several hundred years, marble slabs were systematically stripped from the map and used to construct new buildings, or simply burnt in kilns to make lime for cement [30].

As presented by Adán et al., the process of defining the automatic reconstruction of archaeological pieces through the integration of a set of unknown segments is a highly complex problem, and when only a few segments of the original pieces are available, the solutions, exclusively based on computation algorithms, aim to create credible whole restorations. The authors have proposed the use of hybrid human/computer strategies to tackle incomplete 3D puzzles. The method has been applied successful on the fractured pieces belonging to the remains of Roman sculptures at the well-known Mérida site in Spain [31].

Most of the 3D reconstructions of cultural heritage projects are focused on archaeological pottery. Ceramic studies have played a central role in the development of archaeology, as ceramics represent by far the largest class of artefacts recovered during the excavations of historical sites. A method intended to assist in the tedious task of reconstructing ceramic vessels shards using 3D computer vision has been presented by Cohen and Ezgi [32]; their proposed method aligns the shards based on a set of 3D weighted curve moments; the main advantage of the proposed method is that it can be extendable to surface markings.

For physically reconstructed cultural heritage objects such as ceramics, it is important to capture the photorealistic 3D model. A post-reconstruction of photorealistic 3D models of ceramic artefacts intended for use in interactive virtual exhibitions is presented by Chow and Chan [33].

A recently published review paper by Angelo et al. presents the computer-based methods for the classification and reconstruction of 3D high-density scanned archaeology pottery [34]. The review paper aims to provide a complete and critical analysis of the state-of-the-art until the end of 2021 of the most important published methods on pottery analysis, classification, and 3D reconstruction.

In case of complex cultural heritage objects, 3D laser scanning is combined with 3D modeling techniques to define 3D reconstructions. A case study that combines cultural heritage and ship design is presented by the authors Arapakopoulos et al., for a traditional Greek Boat with Trechadiri hull type named after “Aghia Varvara”; the boat is characterized as a modern cultural heritage monument by the Greek Ministry of Culture [35].

The use of 3D laser scanning and additive technologies for the reconstruction of damaged and destroyed cultural heritage objects is presented by the authors Parfenov et al.; they have been successfully used for outdoor sculptures in St. Petersburg that have been destroyed during WWII [6].

Other researchers have proposed a reconstruction and analysis method for a bronze battle axe, and have compared the inflicted damage injuries using neutron tomography, manufacturing modeling, and X-ray microtomography data [36].

As presented by other researchers, the main risk of a virtual reconstruction is the lack of a faithful reconstruction. Their work proposed a systematic workflow that integrates 2D and 3D sources [37]. Another recently published state-of-the-art approach that has the advantage of making the reconstruction activities easier and less arbitrary is to integrate iconographic 3D comparison elements within the reconstruction process [38].

Implementing 3D reconstruction within digital interactive applications represents an important tool in safeguarding cultural heritage and making it available to future generations, and supporting immersive learning opportunities [39].

3. Materials and Methods

Within this paper, we have proposed a framework (Figure 2) that makes use of an ornament vector displacement map extraction process, which creates a stencil database to enable accurate 3D reconstructions. The proposed framework can be used to acquire, store, and position ornaments as 3D modeling tools to address 3D reconstruction challenges and associated problems (self-intersections, non-manifold vertices, 3D model topology, and file format interoperability), enabling faster digital reconstructions.

The framework has multiple steps, starting with the input fragmented artefact that is processed within the documentation and 3D scanning phase. The novelty of the proposed framework when compared to other recently published 3D reconstruction frameworks is that it makes use of the 3D geometry associated with the 3D scanned and processed ornaments that are stored as vector displacement maps. Other frameworks integrate the 3D scanning directly within the reconstruction process, but this can lead to numerous challenges that have been described within the introduction of the paper. These can become problematic within the dissemination and promotion phase, when computer-aided manufacturing technologies are used to create tangible 3D reconstructed models.

3.1. Documentation and 3D Scanning Phase

The 3D scanning phase has been conducted directly within the National Museum of Transylvanian History from Cluj-Napoca, Romania, using the equipment available within the 3D scan laboratory of Technical University of Cluj-Napoca [40]. Within the laboratory, there are a wide variety of 3D scanning equipment, such as laser scanners, structured light scanners, terrestrial laser scanners, and coordinate measuring machines equipped with 3D scanning accessories, as well as photogrammetry equipment.

Displacement mapping techniques generated from 3D scan datasets have been previously applied to cultural heritage. The idea of coding a dense mesh with fine details and mapping them on a simplified version mesh has been pursued since early computer graphics applications. The researchers Guidi G. et al. have used displacement mapping techniques to optimize mesh models generated with 3D scanners [41].

The proposed framework makes use of the displacement mapping technique, but they are not intended to be used as simplified mesh versions; instead they are used to store the fine details of the 3D scan dataset as stencil tools to facilitate the 3D reconstruction of fragmented cultural heritage objects and monuments.

The digitization process made use of a structured light scanning solution to record both the 3D shapes and the color texture of the Dacian embossed disks at a high resolution. We have opted for 3D scanning with the Creaform Go!Scan 50 [42], as it provides trueness and precision that ensures that the ornaments are acquired at their true scale with minimum deviations. The scanner has been developed by Creaform in partnership with AMETEK, as a flexible and rapid 3D scanning solution that is designed to address a wide range of professional needs, starting from industrial applications up to cultural heritage, while providing a portable handheld solution. The scanner makes use of 3D positioning targets, but it can also be used with ease for various cultural heritage digitization tasks, as it can use both targets, geometry, and texture as positioning references.

The structured light scanner has a large field of view, with a scanning area of 380 × 380 mm, and is recommended for objects with a size of between 300 and 3000 mm, making it the perfect solution, considering the size of the fragmented Dacian embossed disks from Piatra Roșie. The resulting 3D model is presented in Figure 3, along with close-ups of the most important ornaments.

The framework can make use of any 3D scanning equipment that is available, for the case study of the fragmented embossed disk. This was also processed using photogrammetry and laser scanning, but the most accurate and detailed meshes were obtained using the Go!Scan 50.

Having 3D scanned the fragmented embossed disk ornaments, the raw dataset of each ornament has been processed using the following proposed ornament vector displacement map workflow (Figure 4). For this case study, an Autodesk Mudbox software solution was used, as it integrates an operation that enables vector displacement map creation. Other software solutions that enable vector displacement map creation can be used, such as Zbrush, Modo, and even free and open-source Blender. The vector displacement map can be processed within MATLAB. We have chosen Mudbox, as it offers better file interoperability than other similar software solutions. The interoperability from Mudbox is enabled by using Autodesk 3ds Max, which can be linked to Mudbox, offers a wide variety of file formats to be imported, and integrates topology modifiers to ensure that the 3D mesh has no associated 3D modeling problems.

The ornament extraction process is based on ray casting methodology. The displacement map generation is performed within Autodesk Mudbox, and the process requires a source model (high resolution) that contains the sculpted details to be extracted as a reusable displacement map. The source models have been generated using the 3D scanned dataset, and it is important to note that this model does not require UV texture coordinates.

This research made use of the Autodesk Mudbox Ray Casting technique to extract the displacement maps by recording the distance (as a pixel value) between the source and target model, based on ray casting along the surface normal of the target surface. To improve the accuracy of the maps generated based on the 3D scanned ornaments of the fragmented disk, the methodology makes use of the vector displacement map technique. A vector displacement map is a 2D image that records both the height and the directional information of individual points as 32-bit floating-point color information. The file can store the distance at which a vertex will be displaced (based on the pixel value in the image map) and the direction of the vertex displacement.

Vector displacement maps have a main advantage over the ray casting extraction method of not producing any artifacts when applied to another surface.

To extract the details from the 3D model acquired using the structured light scanning solution, we have opted for the technique known in the literature as the “Vector displacement map”. This technique has evolved from the technique known as the “Displacement map”, and was developed in the Pixar animation studio for the RenderMan Software solution to allow the creation of details on the surface of 3D models. Other researchers have proposed algorithms that allow the use of video cards to generate these textures that allow for the movement of the geometry of the 3D model [43].

The resulting 3D scanned model has been imported to 3ds Max to facilitate the extraction of each individual ornament. The authors have proposed the following workflow that can be used to enable the extraction of ornaments, as vector displacement map tools that can be rapidly used to reapply the ornament onto the surfaces of other 3D models.

The extraction and processing of the 3D scanned ornaments methodology proposed by the authors (presented in Figure 3) makes use of a two-step process. The methodology makes use of two 3D computer graphics software (3ds Max and Mudbox) from Autodesk, but the workflow steps can be followed with other software solutions such as, Zbrush, Modo, or other commercial 3D modeling software packages, or free open-source software such as Blender. If other software solutions are used, the methodology requires a few adjustments, but it follows the same steps. For this case study, we have chosen 3ds Max and Mudbox, as the two software are interconnected, and as 3ds Max has good interoperability with computer-aided design software such as CATIA, the software that is used to define the accurate 3D model of the support of the reconstructed embossed disk based on the measured dimensions of the 3D scan dataset.

The 3D model acquired by using structured light scanning was exported in two different file formats to facilitate the ease of work when using these files. The file format OBJ is suited for Digital Content Creation software such as 3ds Max, while the STL file format is more suited for engineering analysis for computer-aided design software solutions such as CATIA V5.

The first ornament extracted is represented by the aurochs, which is located in the central register of the disk. The extraction involved the selection and removal of the 3D scanned mesh that is not associated with the ornament; this step is illustrated in Figure 5.

The aurochs’ ornament has several perforated holes on its surface, which is due to the corrosion process of the iron. The corrosion of the ornament was stopped because of the physical conservation process applied to the fragments when the fragments were arranged on their current support. The ornament is presented in Figure 6, both as a close-up photograph and as the resulting 3D cleaned mesh.

The resulting cleaned mesh of the aurochs’ ornament geometry consists of over 600,000 polygons; to transfer all of the details onto a plane surface in Mudbox, an extraction support was used, and the geometry of the plane was subdivided to level 7. The initial plane prefab tool generated in Mudbox consisted of 100 polygons, and each subdivision stage divided each polygon into four polygons. The seventh stage of subdivision generated a plane defined by 1,638,400 polygons. The overlapping ornament positioned over the plane is illustrated in Figure 7.

The ray casting algorithm was used as the extraction method. In Mudbox and other similar applications, this algorithm sends light rays from the top view; in this case, the image with a resolution of 8192 × 8192 pixels was generated to capture all the details of the model that were obtained using structured light scanning. By calculating the distance that these rays travel until they intersect the surface of the 3D model, the algorithm accurately determines the position of these surfaces and saves them in a 32-bit TIFF file.

The resulting displacement map can be added to Mudbox as a stencil tool to allow the ornament to be positioned on any 3D model. The result of applying the previously generated displacement map using the “Sculpt” function is presented in Figure 8.

The resulting 3D model obtained using the ray casting displacement map has sharp geometry on the entire outline of the aurochs’ ornament. In Figure 9, these areas are presented as zoomed in viewports with a wireframe visualization activated to highlight the precision of the initial scan generated by the structured light scanner.

These areas can be adjusted during post processing by applying the subdivision modifier and local smoothing in the required areas.

The following step involves the digital reconstruction of the areas that were perforated or missing. In the case of the aurochs’ ornament, there were several areas that were missing, such as the ear, the mouth, and other small areas where the surface had been corroded. The process of defining the geometry in these areas involves the use of tools that add geometry, such as “Fill” and “Sculpt” within Mudbox. It is important that this step is done now, after the resulting ray casting displacement map has been generated, so that the resulting mesh is clean and optimal for additional features. These tools can be applied directly onto the 3D scanned meshes, but there is a high chance that the meshes will have elements that are referred to in the literature as “non-manifold”.

The result of the filling process is presented in the figure below; since these areas are completely missing from the model, certain details can be sculpted onto the surface of the ornament so that they merge and blend with the rest of the elements. As is presented in Figure 10, the 3D model contains very fine details on its surface, such as the granulation on the surface of the aurochs that define the fur.

Having the entire 3D model filled, the next step involves the creation of the 3D vector in 32-bit EXR format, which can be used to recreate the ornament with all the missing elements on the surface of any 3D model. This technique is known as “Vector Displacement Map”, the configuration interface is like the ray casting algorithm, as a “Target Model”, a simple plane, and for the “Source Model”, the displaced plane that contains the 3D model of the ornament. For this case study, the same resolution of 8192 × 8192 pixels is used. Due to the very large file size of the resulting file (466 MB), this file format is similar to a picture, but consists of several layers and uses RGB information so that Mudbox and similar applications can recreate these details on any 3D surface. Within Mudbox, there is a dedicated area where users can view these layers—entitled “Image Browser”. The image can be browsed with ease using the keyboard “+” and “–“ keys with a increment value, which can be adjusted from the application. Figure 11 presents two different layers of the resulting EXR file visualized in Mudbox.

Using the same workflow, all the ornaments of the aurochs’ embossed disk have been processed. The 3D scanned model, 3D vector, and the 3D resulting mesh are presented in

Table 1. The extracted ornaments of the aurochs fragmented embossed disk.

3D Scanned Model	EXR Extracted File	3D Resulting Mesh

.

3.2. 3D Digital Reconstruction Phase

As presented within the introduction, the fragmented embossed disk was reconstituted in 1954 incorrectly in four registers, but as other similar embossed disks have been discovered and repatriated in 2011, as well as photographs or other similar shields that have appeared online, it is certain that the Dacian embossed disks only had three registers. The proposed reconstruction from 1954 integrates the fragments on two different disks, one with an aurochs in the central register, and one with a griffin in the central register. The griffin leg and the circular patterned leaves are very similar to the other repatriated shield from 2011 that is currently within the National Museum of History of Romania; these are illustrated in Figure 12.

Within the following section of the proposed research paper, the authors have proposed a reconstruction of the aurochs disk based on previously extracted ornaments. Along with the 3D scanned fragments, the authors made use of the 3D scan model of the repatriated aurochs’ disk from 2011. The similarities between these disks, as well as the other aurochs’ disk, for which there are only photographs that have been listed at an online auction house available, are presented in Figure 13.

To measure the fragmented embossed disk size, the 3D model has been imported to the CATIA V5 software solution. Other CAD software (such as: Inventor, Fusion, or even open-source FreeCAD) can be used to follow the proposed framework. We have chosen CATIA as it provides good surface-based modeling and reverse engineering functionalities. The use of a CAD software capable of importing 3D scans at their true scale, which makes use of parametric modeling, is mandatory for defining an accurate 3D reconstruction proposal. Other software solutions such as Blender or 3ds Max can be used to a certain extent, but they only offer a limited parametric modeling functionality that does not enable complex 3D reconstructions.

The diameter of the central register has a value of 192 mm; the following register has a diameter of 304 mm, and it contains the floral ornaments patterned in a circular array around the central register; the outer diameter has a value of 405 mm. There is a deviation of concentricity of 3 mm between registers number 1 and 2. The outer area of the disk is in an advanced state of degradation; since this area lacks any information, it will be remodeled like the two embossed disks that have been repatriated in 2011, and which are well preserved. The width of this area has a size of 50 mm, and will have 12 perforated holes. The dimensions of the registers measured in CATIA V5 are presented in Figure 14.

The 3D model of the support was modeled considering the positioning of the existing fragments and their layouts. To define the 3D support surface of the disk, the 3D scan of the 2011 repatriated aurochs shield was used as a reference and was colored in green. Figure 15 presents various aspects and a section view between the two 3D scanned disks, which are overlapped to define the profile of the support reference for the proposed digital reconstruction; the support surface define is colored in yellow.

Having the CAD models of the support shield, the next step involves the conversion of the 3D CAD model into a polygonal model to a tessellated mesh to facilitate further subdivision within Mudbox. This step is important, as CATIA software uses a curve-based system to define 3D models, and the geometry is not defined by vertices and polygons.

The first register is delimited by the central register, by a rope-like embossed band. If, in the case of the embossed disk from the National Museum of Romanian History, there are 19 leaves in the case of this shield, 19 leaves cannot be framed. Using the 19 leaves distributed on the diameter of the second register, which has a diameter of 192 mm, would result in an angle of 18.947 degrees between the individual elements of the leaf-type ornament, which would create an overlap. Since six consecutive leaves have been preserved, their number can easily be determined. In total, in the second register, there are 14 leaf-type ornaments distributed at 25.714 degrees between them. Similarly, the number of ornaments positioned between these leaves can be determined, which is also equal to 14 elements also distributed at 25.714 degrees around the inner register, as is presented in Figure 16.

Having the 3D model of the support defined, the vector displacement maps generated, and the correct number of circular array ornaments, the following step involves the processing of the digital reconstruction within Mudbox. One important step regarding the interoperability of 3D models between CAD software such as CATIA and other 3D applications such as Mudbox is that they require different type of file formats.

The resulting 3D surface from CATIA has been opened with 3ds Max software to facilitate the retopology of the surface to be suited with the polygonal mesh type that is compatible with Mudbox software. The Retopology made use of the ReForm algorithm from 3ds Max, and the target face count has been set to 2000. The resulting retopology surface is presented in Figure 17.

The resulting mesh has been transferred to Mudbox with the interoperable send to system implemented within 3ds Max, where the surface has been subdivided to level 6, which means that the surface is defined by almost 2 million vertices. Within 3ds Max, the 3D scanned fragmented embossed disk has also been imported to serve as a reconstruction reference. The 3D scanned file has also been modified using the ProOptimizer modifier to reduce the number of vertices to preserve enough information within the 3D model to serve a reconstruction template; therefore, it is important to have the accurate location of each ornament in relation to the 3D modeled support surface. All the vector displacement maps generated within the previous step for all the ornaments have been added to the software as stencil tools. The overlap between the 3D scan used as a reference and the support of the embossed is illustrated in Figure 18.

Following the position of the 3D scanned model, each previously extracted ornament has been applied on the surface of the disk. This step involves the use of the Sculpt Tool and the previously generated displace vectors. The phase regarding the application of the aurochs’ ornament and the circular ornaments are presented within Figure 19.

All the previous ornaments extracted have been applied on the 3D model using the same workflow. The main advantage is represented by the ease of use regarding the position, rotation, and scale of the stencil, as these can be adjusted rapidly. For each stencil, it is recommended to make use of the camera bookmark option, which will save the camera location. The ornaments should be applied only from the top view, so that they will be accurately positioned on the surface of the proposed 3D reconstruction.

4. Results

4.1. 3D Reconstructed Embossed Aurochs’ Disk

Having the surface of the proposed reconstruction finished in Mudbox, the resulting model was transferred back to 3ds Max so that the final model would have the 12 holes positioned on the exterior register, similar to the other Dacian embossed disks. These have been added using the Boolean operation from 3ds Max onto the resulting mesh. Having the final mesh of the reconstructed disk, the thickness was set to a value of 1.8 mm, which is the average thickness of the two repatriated embossed disks that have been 3D scanned to serve as a digital reference for the 3D reconstruction process. The resulting 3D model is presented in Figure 20.

Within the final step, the 3D model has been textured using various iron textures (hammered iron, damaged iron, and rusted iron) to define various aspects of the visual appearance of the proposed 3D reconstruction model. The 3D textured models have been integrated with the project online database as embedded 3D models from the Sketchfab platform to facilitate their dissemination and interactive visualization. The resulting 3D models have also been saved as individual files to facilitate their integration within the digital application, as well as computer-aided manufacturing.

4.2. Ornament Database with 3D Scans and Vector Displacement Maps

All the ornaments of the 3D scanned Dacian embossed disks that are currently within the inventory of the two Romanian museums (the National Museum of Romanian History and the National Museum of Transylvanian History) have been extracted and added within the 3D Dacian embossed database on the project website. The vector displacement maps extracted using the workflow presented within this research paper have also been added as EXR files hosted on Dropbox platform. The interactive database from the project website is presented in Figure 21. The digital ornaments are organized in collections, one for each embossed disk, which also include the associated initial 3D scan of the embossed disk.

Within the database, for the ornaments and the 3D reconstructions of the two Dacian embossed disks, for which photographs have appeared online at an auction house, the authors have recreated them based on these images paired with the measurements of the three similar embossed disks that have been scanned. One of the embossed disks has an aurochs as the central element, while the other one has a stag. The two 3D reconstructions are presented in Figure 22. These two embossed disks have also been manufactured using an additive manufacturing technique; the workflow is presented within a related research paper published by the authors [44].

4.3. Virtual Reality Application

Virtual reality represents an effective use of technology to visualize and interact with 3D reconstructions, as head mounted display (HMD) systems have become widely available and affordable. The use of immersive and interactive virtual reality systems reduces the distance barrier between the public, cultural heritage sites, and associated buildings and artefacts that often have been either destroyed, deteriorated, relocated, or even missing, as was the case of the looted Dacian ornamental disks repatriated in 2011, or the ones still being searched for by the Interpol.

The developed virtual reality system application is intended for the Oculus Meta Quest 2. This headset is lightweight and represents an all-in-one system solution that can easily be packed and transported to indoor and outdoor museum exhibitions. This represents the perfect solution for outdoor events such as Open Day events at Sarmizegetusa Regia archaeological site, or even at the Dacian Hillfort from Piatra Roșie. Figure 23 presents an aerial view of the Sarmizegetusa Regia Sacred Area and various aspects from the Open Day Event organized directly within the archaeological site, with two large tents and a wide variety of digital equipment and applications.

The virtual reality application was developed using the popular Unity engine, which is a cross-platform game engine capable of deploying a wide variety of digital applications mostly used for game development, but also widely used as the main platform for 3D cultural heritage applications. The solution can develop applications both for computers, consoles, mobile devices, web-based applications, virtual reality systems, augmented reality glasses, and mixed reality glasses.

Since Meta Quest 2 represents a portable all-in-one virtual reality headset based on Android OS, the processing capabilities are limited when compared to a virtual reality headset that is directly connected to either a laptop, a computer, or that makes use of tethering capabilities. Since the application is intended for either indoor museum exhibitions or remote areas such as the Dacian Fortresses of the Orăștiei Mountains archaeological site, where tethering technology is limited or there is no connectivity, the 3D application was highly optimized to integrate low polygon 3D models. The digital environment used within the application represents a 3D reconstruction of the Sarmizegetusa Regia Sacred Area (presented in Figure 24), which can be explored by using the teleporting functionality of the controllers. The Dacian embossed disks have been positioned on pillars, close to one of the temples to facilitate the ease of interaction.

Figure 25 presents three snapshots of the developed virtual reality application. The casting option of the Meta Quest 2 virtual reality system has been used to stream the output to a laptop where the screen has been recorded. The application can be used with ease to pick up, analyze, and interact with all of the five embossed disks that have been either 3D scanned or 3D reconstructed. For the fragmented disk, which represents the case study of this article, we have integrated within the virtual reality application both the current physical reconstruction and the proposed 3D reconstruction.

4.4. Augmented Reality Application

As the research project aims and scope were to reconstruct the fragmented and looted embossed disk and to facilitate their dissemination, one of the best digital dissemination solutions makes use of an augmented reality application that can easily be installed directly on visitors’ smartphones and tablets, which are widely available and commonly used. The application has been developed with Unity and makes use of the popular Vuforia Engine to enable real-time tracking based on images to overlap realistic 3D rendered reconstructions of the embossed disks. An overview of the application is presented in Figure 26.

To extend the functionality of the application by providing additional information regarding the Dacian embossed disks, within the application, we added various video files regarding the 3D scanning process and 3D reconstruction, as well as the metadata associated with each known Dacian embossed disk. The functionality of the application is presented in Figure 27. Currently, the application makes use of the defined image targets, generated using a Matcap shader to highlight the fine details of the Dacian embossed disks.

As a future work, we want to make use of the tangible reconstruction of the disks created using additive manufacturing to enable 3D model tracking. Our final intent with the augmented reality application is to allow visitors to interact with the tangible reconstruction directly, and to have access to enhanced visualization and metadata on demand, based on the augmented reality application. We have started to print the embossed disks at their true scale, but having the object scaled down will be better, as these will be easier to handle with one hand, allowing users to use the other hand to track the object. Figure 28 presents the manufacturing of the tangible embossed disk at its true scale, but also at a smaller scale, on two different 3D printers.

4.5. Mixed Reality Application

The developed virtual reality application enables a better immersive interaction with the 3D reconstruction than the augmented reality applications. Within the project, we have explored the used of mixed reality, which blends the physical and digital worlds. The use of an augmented reality application that makes use of the tangible embossed disk obtained using additive technology can enable better interaction with the 3D reconstruction; the only drawback is that the user must hold the smartphone/tablet so that the tangible embossed disk is tracked, and this usually requires the user to use both hands.

This is where mixed reality glasses such as Hololens 2 can provide a better immersive interaction, by combining and enabling interaction with and among real-world and virtual objects. Figure 29 presents the application developed in Unity, which was built specifically for the popular mixed reality device—the Microsoft Hololens 2 glasses. This equipment can track the environment around the user automatically and in real-time, as is presented below, thus enabling real-world data integration within the virtual environment.

5. Discussion

This research work aims to reconstruct the original shape of a fragmented artefact by enabling a faithful 3D model to be used as the support of the embossed disk modeled within CATIA, a computer-aided software specializing in surface design and reverse engineering. The resulting 3D model has then been transferred to Mudbox to enable the use of the defined ornaments as stencils, which enable the deformation of the 3D modeled support vertices to follow the trueness and accuracy of the 3D scanned ornaments. The ornaments could have been positioned within CAD as overlapping meshes, but we wanted to ensure that the 3D model has no self-intersection and other 3D modeling related problems that would later raise problems in computer-aided manufacturing processes. Existing 3D reconstruction frameworks published in related articles make use of 2D and 3D references to define accurate 3D reconstruction. Our proposed 3D reconstruction framework proposes the creation of an ornament stencil database that eliminates problems regarding the 3D model topology and other associated 3D modeling problems.

The shortcoming of the proposed framework is that the use of detailed vector displacement maps requires a high amount of processing power; therefore, the use of a performance laptop or workstation with at least 32 GB of RAM is recommended. As presented within the paper, the framework can be followed using open-source software such as Blender to replace Mudbox, or FreeCAD to replace CATIA, but there will be some interoperability problems associated with the file format, which will have to be resolved.

The accuracy of the proposed 3D reconstruction is closely linked to the specification of the 3D scanner that has been used to acquire the 3D dataset. The fragments of the disk currently exhibited at the National Museum of Transylvanian History from Cluj-Napoca and the two repatriated embossed disks exhibited at the National Museum of Romanian History from Bucharest have been 3D scanned using the Creaform Go!Scan 50 structured light scanner. The authors made use of this scanning solution as it provides trueness and precision compared to the other 3D scanning equipment currently available within the 3D scan laboratory from the Technical University of Cluj-Napoca. There are other scanning solutions commercially available on the market that can acquire more accurate 3D datasets, but those solution are either too expensive, not portable, or not accessible to us.

The vector displacement maps generated based on the available 3D dataset obtained integrate fine details from the surface of the ornaments. As an example, the features of the aurochs’ ornament have intricate details (the tuft of fur under the snout, the high withers, and the mane). Even the small parallel incisions used to render the fur have been acquired by the scanner and stored within the resulting vector displacement map after the proposed ornament extraction framework.

As for the case study, we have selected an artefact that was physically restored incorrectly by curators soon after it was discovered. The incorrect restoration integrates ornaments from two different Dacian embossed disks. Only with the recent repatriated Dacian embossed disks was it clear that the fragments are from two different disks; therefore, the 3D digital reconstruction process described within this paper has been conducted. Only the aurochs’ embossed disk has been 3D reconstructed, as the other fragments with the griffin leg and the circular patterned leaves are too fragmented to enable an accurate 3D reconstruction. These ornaments have various similarities to the griffin repatriated disk (Figure 12), but the leg position from the central register is different; therefore, the 3D reconstruction proposal would be subjective.

We have reached the proposed 3D reconstruction framework after several other 3D digital reconstruction attempts of the Dacian embossed disks. From a visual point of view, the other reconstruction attempts that made use of overlapping 3D meshes were no different, but the resulting mesh had various self-intersection and non-manifold problems that have started to raise problems within the manufacturing process. As 3D models used for 3D web visualization or digital assets for the VR, AR, and XR environments, 3D meshes that have various interior self-intersections and overlappings do not represent any real concern, as they can be visualized and manipulated without any problems. The problems appear when sectioning the 3D reconstructions within these virtual applications.

The ornaments were embossed on the real artefact by hammering the reverse of the heated iron sheet, and the details were engraved on the obverse side using special instruments. Therefore, the proposed framework based on vector displacement enables the digital reconstruction to be performed as with the real creation method by displacing the existing material to define the ornaments.

6. Conclusions and Future Works

This article describes a framework that can enable an accurate 3D reconstruction of fragmented or damaged cultural heritage assets. The framework can be applied with any 3D dataset acquired using a wide variety of 3D digitization technologies such as (laser scanning, structured light scanning, terrestrial laser scanning, and even less accurate technologies such as photogrammetry).

The framework can be adjusted if other software solutions are used. The authors have chosen Autodesk 3ds Max, Autodesk Mudbox, and Dassault Systèmes CATIA V5, as these have good file format interoperability, allowing 3D CAD models and 3D scans to be added to Mudbox with the required 3D model topology (edge distribution and structure). We made use of CATIA V5 as it provides engineering tools for 3D scanning and reverse engineering, as well as a great surface modeling workbench.

The proposed 3D reconstruction of the fragmented Dacian embossed disk, 3D scan datasets, individual ornaments, and the resulting EXR files are available on the project website. The digital application developed to promote and disseminate the project results and the Dacian embossed disks will be published on the project website. Their aims and scope are to raise awareness and make the Dacian embossed disks known to the public. The authors will prepare videos to help with the dissemination of these digital applications, as video links are easy to be shared by users, to facilitate widespread dissemination.

Aspects relating to open and reusable 3D models are addressed by the 3D database, with 3D scans, reconstructions, and individual ornaments. This leverages digital asset reuse and open access of the research project, allowing other researchers to analyze and to compare similar ornaments.

Another important existing challenge associated with 3D digitized and 3D reconstructed cultural heritage is that they are not linked and not connected with their associated metadata as provided by cultural heritage institutions. Therefore, we have proposed the use of VR, AR, and XR digital applications as dissemination and promotion tools. Within these digital applications, the metadata can be added as text, images, external links, and audio and video files to provide on-demand metadata to the users directly within the 3D interactive environment.

Another important tool that is intended to help with the widespread dissemination and project findings is represented by the augmented reality application, which represents the easiest way to distribute the digital interactive application, as mobile devices capable of processing the application are widely spread, and the targets used by the application can easily be printed or even used digitally if the reference image targets are open on any type of display.

Virtual reality is more popular than it has ever been and as the technology continues to evolve, virtual reality systems become more lightweight and more powerful, and begin to integrate hand tracking gestures, thus reducing the need for specific controllers to interact with and navigate the virtual environment. All-in-one virtual reality systems such as Meta Quest 2 do not need to be connected to a powerful workstation or a laptop, and they are capable of processing detailed virtual reality environments. The application makes use of optimized 3D scans and 3D reconstructions.

Mixed reality devices such as Hololens 2 are currently cost-prohibitive for consumers, but as the technology evolves, these devices will follow the path of virtual reality headsets and will become more affordable. Their current processing power enables the use of 3D scanned and 3D reconstructed objects, but they also require heavy processing and optimization to enable good frame rates when the user wants to interact, grab, scale, or rotate these digital assets.

One future direction of our research involves the use of deviation analysis on the 3D scan datasets to identify if the creation of the Dacian embossed disks involved the same tools used in the hammering of the reverse of the heated iron sheet. This process involves the use of specific reverse engineering tools and techniques to compare the 3D scan datasets.

Furthermore, our aim is to analyze these 3D scanned datasets with other Dacian cultural heritage assets that integrate zoomorphic representation, such as the more recently discovered bronze matrix with hollow designs. The bronze matrix with hollow designs has eight faces: the main two are hexagonal and the other six are rectangular, and they integrate a wide variety of zoomorphic representations, including aurochs and griffins, which are also in the central register of the Dacian embossed disks. The bronze matrix also includes other zoomorphic representations, such as: lion, tiger, leopard, elephant, rhinoceros, hippopotamus, bear, boar, wolf, bull, wisent, dog, deer, horse, goat, antelope, and rabbit. The theme of the scenes present on the matrix (fights between animals) is very old and widespread over vast cultural spaces.

Another future direction involves the 3D reconstruction of the building where the fragments of the embossed disks have been found, and to create a virtual reality environment based on terrestrial laser scanning of the area. Within this virtual reality environment, the Dacian embossed disks will be exposed with the help of beautifully decorated ornamental targets. The griffin Dacian embossed disks repatriated in 2011 have a decorated ornamental target on the top left corner (Figure 12).

The virtual reality, augmented reality, and mixed reality applications will be further improved to integrate additional features. Currently within the application, the metadata are added as text, images, audio, and video files that present various aspects regarding the 3D scanning and 3D digital reconstructions. We are considering adding metadata for additional languages within the application; currently the applications are in English, as this is one of the most widely used and popular languages in the world.

To define the tangible replicas, both at their true scale and as a target object for the augmented reality application, we made use of fused deposition modeling, using ABS, as these are currently commonly available and affordable. As additive manufacturing has evolved, there are technologies that make use of direct metal laser sintering, which has become more affordable. We also consider analyzing the resulting metal additive manufacturing replica with a replica generated using reductive computer-aided manufacturing.

As stated by Marc-André Renold [45] “The very lucrative black market in works of art antiques has flourished thanks to the keen interest of buyers, shortcomings in legislation, the complicity of those in the sector, an increase in looting in countries in conflict situations and the development of online sales platforms.”

We consider that raising awareness can have a positive impact on ensuring cultural heritage preservation. Hopefully, the two Dacian embossed disks known to be looted will be found and repatriated, as Cultural Heritage is a non-renewable resource for a nation.