Abstract
This paper presents the results of the 12-month project "Seals and Their Impressions in the Ancient Near East" (SIANE), a collaborative effort between the University of Southampton, Oxford University, and the University of Paris (Nanterre). Recognizing the limitations of traditional visual documentation for cylinder seals, SIANE developed a portable structured light scanning system and workflow designed for high-throughput 3D digitization. By capturing high-resolution geometry and generating automated "digital unwrappings," the project facilitates corpus-wide research, including computer-aided image recognition. This methodology offers a potential solution for linking original seals to ancient impressions on tablets dispersed across global collections.
Introduction: The Visualization Challenge
Ancient Near Eastern cylinder seals served as critical administrative tools and artistic expressions for over 3,000 years. However, their cylindrical form presents a unique documentation challenge. Traditional methods—such as 2D photography or physical rollouts in plasticine—often fail to capture the stone's texture, color, and the precise 3D depth of the engravings. Furthermore, with an estimated 50,000 seals scattered globally, manually matching a seal to its ancient impression on a clay tablet is a significant logistical hurdle. The SIANE project sought to overcome these barriers through a bespoke 3D imaging workflow.
Methodology: The SIANE Scanning System
The project developed a portable scanning rig designed for high-throughput digitization. The hardware consists of a FLIR Grasshopper industrial camera and an LED-based projector mounted on linear stages, with the seal placed on a high-precision motorized turntable.
Structured Light and Digital Unwrapping
Instead of complex laser patterns, the system utilizes a "shadow edge" structured light approach. A simple shadow edge is projected parallel to the seal’s axis; as the seal rotates, the camera records the distortion of this contour to calculate surface geometry. Data processing involves a modified Canny surface detector and iterative Newton’s method for sub-pixel accuracy.
Beyond 3D geometry, the system produces "digital unwrappings." By splicing strips from approximately 120 images taken under general illumination, the software creates a flat 2D representation that retains the object's original color and texture. This digital surrogate allows researchers to view the entire design simultaneously without the mirror-image reversal inherent in traditional clay rollouts.
Computer Vision Integration
The project integrated software from the "Seebibyte" project, which employs a "visual words" approach. This allows for the computational matching of seals to their impressions on tablets by decomposing images into prototypical patterns, effectively removing human bias from the identification process and enabling comparative glyptic studies at scale.
Results and Pilot Study
The SIANE system was tested across several major collections, including the Bibliothèque nationale de France (~850 seals), the Ashmolean Museum (~940 seals), and the Cambridge Museum of Archaeology and Anthropology (~150 seals). In total, over 1,600 seals were imaged, with data hosted on the Cuneiform Digital Library Initiative (CDLI).
Key performance metrics include:
- Resolution: A spatial resolution of 0.0156 mm, surpassing the initial target of 0.02 mm.
- Depth Precision: Accuracy better than 0.01 mm, sufficient for capturing shallow tool marks and engraving details.
- Efficiency: An average scan time of 10 minutes per seal, including handling and focusing.
- Automated Matching: The pilot successfully identified matching seal impressions across different tablets in the CDLI database, proving the viability of automated indexing.
Discussion: Insights and Limitations
A core philosophy of the SIANE project is the prioritization of geometric data over color. While color is useful for presentation, the 3D geometry reveals the "hand" of the ancient engraver through tool mark analysis. Furthermore, the project emphasizes standardization; by fixing the resolution at 0.02 mm, the data remains comparable across different collections and suitable for machine learning applications.
However, technical challenges remain. The narrow depth of field (~1 mm) required to keep exposure times low and avoid motion blur makes scanning irregular or "lumpy" seals difficult. Additionally, seals longer than 50 mm currently require multiple passes and manual splicing. Future iterations aim to automate focusing and alignment to further reduce the manual bottleneck and manage the significant data volume (approx. 4 GB per seal).
Conclusion
The SIANE project demonstrates that a robust, portable, and relatively simple structured light system can revolutionize the study of ancient glyptics. By providing a standardized, high-resolution digital corpus, the project lays the groundwork for automated large-scale research that was previously impossible, offering a path toward digitizing the entire global heritage of cylinder seals within a few years.
Summary updated