3D Digitization of a Composite Medieval Sword: Challenges Between Dry Versus Wet Surfaces

3D Digitization of a Composite Medieval Sword: Challenges Between Dry Versus Wet Surfaces Papoutsaki Roza1, Makris Dimitrios*2, Petrou Maria3, Argyropoulos Vassilike4, Giannoulaki Maria5, Malea Ekaterini6, Rapti Stavroula7. 1 [email protected] Department of Conservation of Antiquities and Works of Art, University of West Attica, 12243 Egaleo, Athens, Greece 2 [email protected] Department of Conservation of Antiquities and Works of Art, University of West Attica, 12243 Egaleo, Athens, Greece 3 [email protected] Department of Conservation of Antiquities and Works of Art, University of West Attica, 12243 Egaleo, Athens, Greece 4 [email protected] Department of Conservation of Antiquities and Works of Art, University of West Attica, 12243 Egaleo, Athens, Greece 5 [email protected] Department of Conservation of Antiquities and Works of Art, University of West Attica, 12243 Egaleo, Athens, Greece 6 [email protected] Department of Conservation of Antiquities and Works of Art, University of West Attica, 12243 Egaleo, Athens, Greece 7 [email protected] Department of Conservation of Antiquities and Works of Art, University of West Attica, 12243 Egaleo, Athens, Greece *Corresponding author Correspondence: [email protected] ABSTRACT This study presents a composite waterlogged archaeological sword, preserved in two parts (A’ and B’), along with the context of the burial environment, as well as the location and the extent of each of the different construction materials. A comprehensive description is provided of the methodology and the technological equipment used for the digital imaging of the examined sword. The results of the digital imaging process are then presented. Imaging of part A’ was successful, while part B’ could not be fully captured due to limitations in the software and the glossy surface caused by the object's wet condition. An attempt was therefore made to digitally join the individual parts in order to present the whole sword as a complete artefact. Although the resulting digital imaging did not present the sword as a unified model whole, it enabled the visualization of the assembly of the two parts. The digitised outputs are positioned both as research assets and as decision-support surrogates for preventive conservation, including how virtual models can inform desalination, corrosion inhibition, and wood stabilization strategies. Keywords: cultural heritage, 3D digitisation, preventive conservation, digital preservation, photogrammetry, composite object Introduction The digitization of archaeological artifacts has become a cornerstone of cultural heritage research, allowing for the preservation, analysis, and dissemination of material remains. Recent advances in three-dimensional (3D) digital technologies provide archaeologists with unprecedented opportunities to study fragile or endangered objects without risking their integrity (Hostettler et al., 2024). However, challenges arise when dealing with composite artifacts or those preserved in waterlogged contexts. In particular, surface conditions—whether dry or wet—can significantly affect the quality of 3D data acquisition (Gil-Melitón, etal. 2019; Stelzner et al. 2022). Between 2007 and 2013, underwater surveys conducted in the port of Rhodes island, Greece, by the campaign of the Ephorate of Underwater Antiquities led by the archeologist Dr. G. Koutsouflakis, resulted in the recovery of four heavily concreted swords and a dagger. Measuring approximately one meter in length, the swords were made of iron combined with various organic materials. They were discovered in close proximity to a 13th-century shipwreck and are believed to have originally belonged to the Knights of Rhodes (the Hospitallers) under the Order of St. John. The sword of the present study (sword 18) is one of these artefacts (Argyropoulos, 2016). Sword 18, is a piece of medieval weaponry, dated to the 13th century AD, and it falls into the "sword of war" category. The sword has a total length of 89 cm. It consists of two main components: the double-edged blade and the hilt. The blade measures 76.5 cm and includes a fuller (central groove) extending for 44.7 cm. The blade is forged from iron, while the hilt is made of wood, constructed around the tang as its core. The sword was recovered within its scabbard, which is made of wood and lined with leather as a chappe. Beneath the guard area, the sword features decorative or functional fittings made from copper alloys (including tin bronze). At the end of the hilt is a pommel, which serves to counterbalance the weight of the blade. The guard, located at the junction between the blade and the hilt, is designed to prevent the hand from slipping onto the blade during use. Below the guard, there is an attachment ring, added as a separate fitting. The sword was recovered in two fragments (Part A and Part B). The sword was brought to the laboratory of Metallic objects of the Department of Conservation of Antiquities and Works of Art, University of West Attica for conservation treatment. It was received in a waterlogged condition and was kept wet throughout documentation and treatment until the final drying process. The condition assessment revealed the sword's composite nature, indicating the presence of iron, copper, tin, and remnants of organic materials from the scabbard. Although metal was identified in several locations of the sword, there were parts in the blades that were completely or partially mineralized. Similarly, the organic materials, leather and wood located on the scabbard, were hardened and were considered mineralized as opposed to waterlogged. It was soon identified that the surface of the sword was heavily contaminated with iron corrosion products, regardless if they are organic materials (wood, leather) or copper alloy inlays. Thus, the surface exhibited a complex stratigraphy, characterized by heterogeneous corrosion layers and mineral deposits (Argyropoulos, et al. 2025). This intricate composition of the artefact resulted in reduced flexibility, increased stiffness, and a heightened susceptibility to brittle fracture. This condition assessment posed significant challenges and highlighted the need of a tailored yet demanding conservation strategy, which was critical for preserving the integrity of the composite artefact. Due to the aforementioned complexities, the two fragments of the sword (Part A and Part B) were not subjected to conservation treatments simultaneously. Part A has already undergone conservation treatment, with both the drying procedure and subsequent surface conservation successfully completed. The drying methodology employed for Part A of the sword prioritized controlled air drying to mitigate risks associated with sudden shrinkage of organic materials. Figure 1 Sword 18, consisting of two parts, labeled A and B. ©Authors However, the condition of Part B was markedly poorer, characterized by extensive fractures and corrosion products. The need for tailored conservation strategies became evident, as the differential preservation of materials introduced significant complexities. Therefore, while the drying methodology for Part B is still under consideration and development, due to its critical condition and the difficulty to deal with a composite object, this project focuses on the digital documentation and restoration of Sword 18’s Parts A and B. Furthermore, the digital representation will be used to compare the drying methodologies in terms of surface alterations afterwards. The aim is twofold, first, to document the current condition of the two parts of the sword, while preserving the sword’s information, form, textures, damage, and materials through digital technologies, such as the creation of an accurate 3D model. This will aid the conservators to have a baseline information during the implementation of drying methodology of part B. Second, provide a 3D digital assembly of the two parts of the sword. The findings contribute to the development of standardized workflows for cultural heritage digitization. Theoretical and Applied Framework Digital Surrogacy and Theoretical Perspectives The digitisation of cultural heritage artifacts has been widely discussed not only as a technical process but also as a theoretical and epistemological issue (Muenster, S. (2022). Digital surrogates—3D models that represent tangible cultural objects— occupy a dual role: they act as documentation records and, increasingly, as scientific objects that inform conservation, research, and interpretation (Cameron & Kenderdine, 2007). Within this debate, questions of authenticity, accuracy, and integrity are crucial. The London Charter (Denard, 2012) and the Seville Principles (Bendicho, 2013) emphasize that digital reconstructions must be transparent about their methodological limitations, clearly documented, and contextualized within heritage knowledge frameworks. When placing the present study within these theoretical debates, it is highlighted that the 3D models of Sword 18 are not merely visualizations, but cultural data, whose legitimacy depends on rigorous methodological disclosure and validation. The case of Sword 18 illustrates these issues with particular acuity. Its composite nature—metallic elements combined with organic scabbard remains—creates complicated surface conditions (corrosion, waterlogging, organic degradation) that test the boundaries of photogrammetric reconstruction. Presenting the sword digitization difficulties, not only establishes a practical workflow but also addresses the theoretical issue of how digital surrogates can effectively represent fragile and incomplete heritage objects. Applied Approaches in 3D Digitisation of Composite Archaeological Objects Applied research on the 3D digitisation of archaeological artifacts has consistently highlighted the challenges posed by reflective, glossy, or degraded surfaces. Early work in photogrammetry for cultural heritage (Baltsavias et al., 1996) established the importance of surface texture for tie-point generation, while more recent studies (Stelzner et al., 2022) demonstrate that composite or waterlogged materials often require hybrid imaging approaches. Structured-light scanning has proven effective for metallic surfaces but struggles with transparent or absorbent organic matter, while photogrammetry provides flexibility and scalability but is sensitive to uniform, lowtexture surfaces. The applied literature therefore increasingly recommends multimodal workflows, where photogrammetry is combined with structured-light scanning, reflectance transformation imaging (RTI), or even computed tomography (CT) to overcome object-specific limitations (Frank et al., 2021). This applied context provides critical comparative value for the present study. The fragmentation of the workflow for Sword 18 Part B into separate acquisition groups (sides and edges) echoes failure modes reported in earlier case studies where insufficient overlap or inconsistent illumination impeded alignment (Chiabrando et al., 2015). Similarly, the reflective qualities of wet corrosion layers correspond to welldocumented difficulties in photogrammetric feature extraction (Guidi et al., 2017). By framing our results within this applied literature, the study demonstrates continuity with established knowledge while also advancing it by offering detailed observations of workflow breakdowns specific to a composite, waterlogged medieval sword. Relevant Case Studies The study by Montusiewicz et al. (2021) investigates the application of structuredlight 3D scanning (3D SLS) to historical garments, using the Emir of Bukhara's costume from 1895 as a case study. Housed at the Samarkand State Integrated Historical-Architectural and Art Museum-Reserve in Uzbekistan, the costume comprises a gold-embroidered gown, turban (salla), and high and low shoes (mahsi and kaushi). These items were selected due to their varying textures and materials, allowing for the evaluation of the method across diverse surfaces. Scanning was conducted using Artec Eva scanners and a Nikon D7200 digital camera, with data processing utilizing Artec Studio, Blender, and MeshLab software. The outcome was 30 point clouds totaling over 6 GB and more than 146 million points, accurately capturing the geometry and color of the objects. The study concludes that the proposed methodology can serve as a standard for the digital documentation of historical garments in museum settings (Montusiewicz et al., 2021). The study by Russo et al. (2023) investigates the 3D digital acquisition of the Piffetti Library, an exceptional wooden interior designed by Pietro Piffetti between 1735 and 1740 for King Charles Emmanuel III. Crafted from poplar, rosewood, olive, boxwood, and holm oak, and enriched with ivory inlays, the library represents a masterpiece of 18th-century Italian decorative art. Now preserved at the Palazzo del Quirinale in Rome, it was documented through integrated 3D surveying methods, combining laser scanning (Focus 3D, Faro) and photogrammetry. Researchers faced challenges due to low lighting and the geometric complexity of the carved surfaces, experimenting with various illumination configurations and data-quality verification tools. Photogrammetry proved to be the most effective technique, providing precise color and texture rendering. The study proposes this workflow as a reference protocol for future digital documentation of similarly complex cultural heritage artifacts (Russo et al., 2023). Equally significant is the integration of 3D digitisation within preventive conservation theory. Preventive conservation emphasizes non-interventive strategies—monitoring, risk assessment, and environmental control—designed to safeguard material integrity before invasive treatments become necessary (Sease, 1998; Manzuch, 2017). Within this paradigm, digital surrogates extend beyond documentation to function as diagnostic and monitoring tools. High-resolution 3D models of the Sword 18 could, for instance, reveal micro-deformations in wood after controlled drying, identify corrosion fronts for targeted inhibitor application, and quantify surface loss before and after desalination procedures. Sword 18 exemplifies the relevance of this integration. The conservation challenges posed by its iron blade (salt-induced corrosion), coppertin components, and organic scabbard remnants demand carefully staged interventions. In this case, 3D documentation can support conservators by providing baseline measurements, enabling temporal comparisons, and facilitating non-contact analysis of zones most at risk. Such applications directly align with the theoretical shift in conservation from reactive treatment to proactive, data-driven risk management (Ashley-Smith, 1999). By linking imaging outcomes to conservation strategies, this study positions digitisation not as a parallel activity but as a core instrument within the broader conservation science framework. Materials and Methods Digital documentation was achieved through photogrammetry, selected for its adaptability to the sword's intricate surface features. For this reason, an improvised photo studio was set up in a dark room at UNIWA. Acquisition Protocol for Composite Archaeological Artifacts During the preparation step, the object was carefully handled using polyethylene supports to avoid direct contact with skin or hard surfaces. Three ethafoam blocks with central notches were employed to support the sword, thereby preventing stress from its own weight. Environmental conditions, including temperature, relative humidity, and lighting lux levels, were recorded. The camera was mounted on a non-slip Manfrotto 055 tripod, and it was ensured that the object’s surface was stabilized, with no active dripping water present. For the camera setup, a Sony ILCE-7RM3A DSLR equipped with a 35mm prime lens was used. The aperture was set between f/11 and f/16, ISO between 100 and 200, and shutter speed adjusted according to the lighting conditions. Manual focus was set on the midline of the object, and in-body stabilization was disabled. All images were captured in RAW format, and reference shots included an X-Rite ColorChecker for color calibration. During the lighting step, two polarized LED panels were positioned at approximately 45° angles, each equipped with two 200-Watt cold-light fluorescent lamps. A linear polarizer was attached to the camera lens and rotated to cancel specular highlights. Lighting was further diffused using softboxes or a light tent to ensure even illumination. For materials with high contrast, such as metal versus wood, bracketed exposures at −1, 0, and +1 EV were captured Figure 2 Design representation of the studio with the two different shooting angles, the lighting, and the placement of the object on the turntable. ©Authors The capture strategy involved placing the sword components on a Falcon Eyes Mini Turntable T360-A3, which has a 60cm diameter and a 40kg load capacity, with scale bars visible in multiple views. A full 360° ring of images was acquired every 10–15° at three elevation angles (low, mid, and high). Additional close-ups were taken of corrosion zones, fractures, and organic inserts. Image overlap was maintained at a minimum of 70%, and data organization avoided splitting images into separate disjoint folders unless necessary. In total, six shooting cycles were conducted for Parts A and B, resulting in 770 images per part with a combined size of over 198 GB. The process resulted towards the appropriate 3D meshes in effectively representing the object's geometry, color, and texture and required over 20 hours to complete. Figure 3 The photo acquisition process is the most crucial step in photogrammetry. It involves taking multiple overlapping images (around 70% overlap) from different angles and heights, while maintaining constant camera settings. The turntable was rotated approximately 10-15° per shot to complete a full 360° rotation. ©Authors Digital Data Management All data generated in this study will be preserved following international standards for digital cultural heritage. Raw image datasets (TIFF/RAW) and derived 3D outputs (point clouds, meshes, textured models) will be archived in sustainable formats. The archival workflow is aligned with the FAIR data principles (Findable, Accessible, Interoperable, Reusable) (Wilkinson et al, 2016) and the OAIS reference model, ensuring that data packages (Submission Information Packages, AIPs) include checksums and documentation for future migration. Each processing stage (raw → aligned → dense cloud → mesh → textured mesh) will be versioned and stored separately, ensuring full reproducibility and transparency for future users This ensures that the sword’s digital surrogates remain both technically validated and scientifically usable for conservation monitoring, scholarly research, and longterm heritage stewardship, (The Chartered Institute for Archaeologists, 2020). For the best possible rendering and the accurate representation of the object’s true colors, it was deemed necessary to convert all photographs to the same color profile using the following software programs: Adobe DNG Converter, ColorChecker Camera Calibration, and RawTherapee 5.9. Images Processing Workflow The image-processing procedure was divided into three stages. In the first stage, the images were categorized into folders according to each side of the sword. Two folders were created for each side and two for each edge of the sword, covering both Part A and Part B. Additionally, for Part A, two extra folders were created to contain close-up shots of each side. Each subfolder, as with the general folder of the sword, included the image containing the ColorChecker Passport. The second stage focused on the creation of a color profile. The image containing the ColorChecker Passport for Part A of the sword was imported into the DNG Converter software to convert and export the file from a .raw format into a .dng file. Subsequently, the ColorChecker Passport image from the converted .dng file was imported into the ColorChecker Camera Calibration software. The software automatically generated predefined bounding areas over the colors of the ColorChecker Passport. Once these bounding areas were correctly adjusted to be approximately centered on each color patch, a new profile was created. The resulting .dcp (digital color profile) file was then saved in the general folder corresponding to Part A of the sword. In the third stage, the created color profile was applied to each image. In the RawTherapee software, each subfolder was imported separately to perform color correction and calibration of the photographs using the previously generated .dcp file. 3D Reconstruction and Digitally Joining Fragmented Parts Following the photographic acquisition and processing phase, the images were imported into a structure-from-motion software (RealityCapture). The method employed involved the automated creation of masks for each photograph of the object. Mask generation facilitated the handling of images from multiple sets covering different parts of the object, as well as the automatic removal of the surrounding background. The workflow for the creation of the three-dimensional model began with the import and alignment of the first set of photographs, followed by a preview of the mesh reconstruction for that set. Masks were then exported for the first set of photographs via the Reconstruction tab (Export → Depth and Mask) into the folder corresponding to each image. This procedure was repeated for all subsequent sets, with six sets processed for Part A and six sets for Part B of the sword. A new project was subsequently created incorporating all photographs and their corresponding masks, and alignment was performed. The ground plane and reconstruction region were then defined, followed by mesh reconstruction and texturing. For optimal texture representation, the final settings were configured as follows: Unwrap settings with Style set to “Fix texel size” for enhanced texture quality, Large triangle removal threshold set to 10,000, and then the Unwrap and Texture functions applied. Finally, the reconstructed mesh model was exported in both .obj and .ply formats, enabling further processing in Meshlab, an editing software (Cignoni et al., 2008) for the subsequent merging of the two parts of the sword. Despite numerous attempts for the Part B of the sword, a complete merge of the image groups — as achieved with Part A — was partially successful. It was decided to create a 3D model from two of the four aligned image groups, for the purpose of presenting the joined components of the sword. Furthermore, it was decided to create two partial 3D models (out of four groups) for presentation purposes — specifically the side 1 and 2 of the sword. While not a fully unified model, this allowed visualizing the connection between the two parts of the sword and facilitated their realignment and joining through precise edge matching. Digital Reassembly Starting with the digitised parts A and B of the sword, morphological and shape considerations were taken into account to proceed with the digital restoration and assembly. In sword fabrication, the blade is characteristically symmetrical about its longitudinal axis, allowing for even mechanical properties on both faces of a double‑edged weapon (Williams 1977). Therefore, an axiality-based approach was pursued for the digital assembly process to provide a complete digital model of Sword 18. In detail, the steps were followed to develop an optimally designed and implemented reassembly method for the damaged sword, a. alignment, b. principal orientation, c. local geometric analysis and d. contour matching. The sword parts were imported individually into a mesh editing software, that of Meshlab, where their real-world dimensions adjusted. A new project was then initiated to properly align and merge the parts, ensuring precise positioning. This methodical approach highlights the careful attention to detail in the reassembling process, resulting in an integrated model that accurately represents the complete sword. In particular, after the creation of the three-dimensional model in RealityCapture, the .ply files corresponding to the two parts of the sword were imported into Meshlab to perform their alignment and merging. Initially, each .ply file was imported separately for Part A and Part B, and the two models scale was adjusted to reflect the actual dimensions of the sword. Following this adjustment, each 3D mesh was exported again as a .ply file. Subsequently, both parts’ .ply files were imported into a new Meshlab project to execute the aligning and merging process. The positions of the two models were carefully adjusted utilising the affine transformation Manipulator tool, allowing translation and rotation along the X, Y, and Z axes to accurately reconstruct the original configuration of the sword. Results In the three-dimensional modeling of Part A of the sword, the results were highly satisfactory. Details such as wear, cracks, and color variations were clearly visible, as well as the accurate rendering of the texture. F i g u r e S E The completion of the 3D model for Part B was hindered by several challenges. First, the method of image capture posed difficulties, as photographing the sword from every angle, including its edges, along with its poor surface features, made the merging process significantly more complex for the structure-from-motion procedure. Second, accordingly, the procedure struggled to merge the aligned image sets due to the sword’s surface complexity, characterized by its dark color, texture, and featureless surfaces. The presence of iron corrosion products created a uniform color and texture, lacking high contrast, which further complicated the software’s ability to differentiate between distinct points on the surface. Figure 4 Sword 18, part A, 3d model results after applying texture in Reality Capture software, side I (up) and ii (down) ©Authors