Documentation of ceramics is a main task in archaeology, because ceramics are the most common finds, used and produced in large numbers by humans for several thousands of years. Archaeologists use analysis of ceramics (Leute,1987) on a daily basis to reveal information about the age, trading relations, advancements in technology, art, politics, religion and many other details of ancient cultures.
Therefore we are developing an automated system for ceramics documentation to help archaeologists to document their finds in an efficient and accurate way, one which can be used for further (computerized) research. The basis of documentation of ceramics is a manually drawn horizontal intersection which is called a profile line (Leute,1987). Furthermore the profile line is the longest elongation around – or cross-section through – the wall of a ceramic defined by the rotational axis (also called axis of symmetry). The term rotational axis relates to the fact that rotational wheels (plates) have been used for thousands of years for manufacturing ceramics.
As the ceramics are found in tens of thousands at almost every archaeological excavation, these drawings require a lot of time, skill and manpower of experts. Consequently our method can also be applied to ceramics found within museums as they typically store and show the most important objects found during excavations all over the world. Therefore we are assisting archaeologists in interdisciplinary projects (Kampel 1999, Cosmas 2001) by using an automated system for acquisition and documentation of ceramics using a 3D-scanner based on the principle of structured light (D ePiero 1996, Liska 1999).
In parallel we are developing a second system for analysis of medieval paintings. Therefore we use pattern recognition techniques to detect and classify brush strokes of under-drawings, which require the estimation of color pigments to “see” through layers of paint.
We were asked for a state-of-the art documentation of a sample of ceramic vessels, today stored in the collection of Greek and Roman antiquities in the Kunsthistorisches Museum Vienna (KHM), including Attic red-figured pottery and white ground Lekythoi (polychrome pottery, Koch-Brinkmann 1999). The aim was a full documentation and a comprehensive description of original and restored parts of the vessels.
Therefore we acquired 3D-Models of 128 red-figured vessels of various shapes in the Viennese collection. Especially for the polychrome Lekythoi in the same collection, we chose to combine the experiences of ceramic documentation on archaeological excavations and the color pigment acquisition and classification for medieval paintings. This had to be done because related methods (e.g. Wehgartner 1983) required the removal of a sample of the surface to determine the color pigments, while our method is based on contact-free measurements using a Spectrometer.
The following sections describe the 3D-acquisition, the multi-spectral readings and their combination (registration) followed by results showing the profile line and top-views for the Corpus Vasorum Antiquorum (CVA) and estimation of volumes and symmetry for future work. Finally, a summary is given.
3D-Acquisition and Processing
For acquisition we use a Konica-Minolta 3D-scanner based on the principle of structured light (Tosovic 2002) having a resolution of <0.1mm, which meets the requirements given by archaeologists for their documentation. Documentation using a 3D-scanner for ceramics consists of several steps. First the main part of the ceramic is acquired using a turntable to acquire the side views. Therefore typically 4 to 6 3D-images depending on the complexity of the ceramics are acquired. Than the bottom and the orifice of the ceramic is acquired using typically 4 to 8 3D-images. Finally the 3D-images are registered and background objects supporting the ceramics are removed (Mara 2006a). The result is a 3D-model describing the surface of the ceramic. To estimate an archaeological documentation for publication and further analysis, we have to estimate a profile-line. This is the longest elongation around – or cross-section through – the wall of a ceramic defined by the rotational axis (also called axis of symmetry). The term rotational axis relates to the fact that rotational wheels (plates) have been used for thousands of years for manufacturing ceramics. This assumption can be made for all ceramics of the ancient daily life and almost every other type of ceramic. Therefore we base our work on using the rotational axis to orient a ceramic (or its fragment) to estimate the profile line as it is done manually by drawings. Therefore we use the principle of fitting circle templates (Gander 1994), which was inspired by manual documentation. Furthermore the creation of such manual drawings is a time-consuming task requiring expert-skills and our system will help to dramatically reduce the time for documentation. Figure 1 shows the use of the 3D-scanner at the museum for history of arts, a photograph and a manual drawn profile-line of a Lekythoi.
This section shows the acquisition of the 105 multi-spectral readings of 17 Lekythoi and vessels having a surface with similar painted surface. The surfaces of these vessels have been measured on a Perkin-Elmer Lamda 900 UV-VIS-NIR (UltraViolett, VISible Light, Near InfraRed) Spectrometer applying the diffuse reflectance technique using a 60mm integration sphere. The reflectograms have been measured from 190nm up to 2500nm in 10nm steps. Initial experiments regarding ceramics showed that decreasing the step width below 10nm dramatically increases the acquisition time, while – due to noise – no gain of information is gathered. The data have been rationed versus a pure SpectralonTM background.
The surface of the vessels are placed near the slit as close as possible, in order to reduce the loss of reflected radiation. A black box was put over the object in order to eliminate external radiation. The whole setup is depicted in Figure 4.
The chosen method has some limitations concerning the evaluation of the intensity of the measurements:
- Due to this restricted setup and the varying size of the objects, the distance between object and slit could not be kept constant. To compensate for different distances, we captured the 3D scene of the acquisition setup, which allows reconstructing the original distance. The further the distance of the objects, the fewer is the intensity of the reflected beam.
- Since the objects are not flat, different curvatures also result in a variation of the measured intensities.
- Some limitations are given due to a fixed size of the measurement area given by the slit of the Spectrometer. So in some cases the pigmented area is smaller than the measurement area, which results in a measurement of a mixture of background and pigment.
Therefore these limitations do not allow an exact measurement of the absolute reflectance. Still the measured curves can be evaluated, since the relative reflectances at a different wavelength can be used to identify pigment specific characteristics without any post-processing. As this is not satisfactory for future analysis and especially not for automated analysis we decided to determine the geometry of the setup by 3D-acquisition of the setup for each multi-spectral reading of every vessel. Knowing the setups geometry is important to estimate the measurements error introduced by the distance between the object and its curvature, because the Spectrometer assumes the distance and the curvature as zero due to the deviation of the beam. The reflectance and the deviation of the beam at a typically measurement distance for our work is shown in Figure 3.2. Furthermore the Figure shows as 5x8mm rectangle of the ideal measurement area (at zero distance) in comparison to the spot we have to expect for readings of curved vessels.
Additionally we have to consider that measurement area integrates all the reflectance of the 5x8 mm rectangle, which generally can include any other color of the paintings influencing our reading. Therefore we have to register and estimate the location of the measurement area on the 3D-models.
Registration of the Multi-Spectral Readings within the 3D-Models
To register (combine) the spot measurement of the Spectrometer we used the same techniques and the same 3D-scanner as for the acquisition of the 3D-models. Having the 3D-models of the Lekythoi as part of the 128 piece collection already available, we needed only one 3D-scan per multi-spectral reading. The setup is shown in Figure 4 and described in detail below.
The following steps are performed for determining the geometry of the scene (see Figure 4):
- Generation of a 3D model of the complete vessel:
- 3D-Acquisition of the complete vessel from different point of views.
- Building a 3D model from these individual model parts.
- 3D-Acquisition of the “beam” and object at measurement position. The beams position was estimated by placing an elongated calibration object with prismatic shape into the beam, and marking the position of the Spectrometer slit. The scanner setup must not be changed between the acquisition of the beam and the acquisition of the measuring scene.
- Registration of vessel model with vessel and beam, calculate the intersection between beam and object in order to determine the measurement position.
As a 3D-scan is performed within less than 30 seconds and the fact that a multi-spectral reading requires up to 10 minutes, we could gain an accurate (<1mm) location of the multi-spectral reading on the vessels surfaces without any noticeable increase of working time for multi-spectral readings solely.
Results – Final Documentation
The final documentation consists of two parts. The first part are based only on the 3D-models and immediately used for publication within a new volume of the CVA. Due to the limits of publication costs the CVA will show “only” profile lines, top-views and unwrapping of the vessels and not the multi-spectral readings. The results of the multi-spectral readings will be published as technical report. The next sub-sections show examples of the final documentation and preliminary results of our ongoing project.
As shown in the first two sections, the fundamental archaeological documentation of ceramics is the profile line. Figure 6a shows the automatically estimated profile line of a typical vessel (KHM IV 350) and intermediate results of semi-automated post-processing. This post-processing consists mainly of addition of the inner part of the profile line as an optical 3D-scanner can acquire only what it “sees”. Furthermore artistic features are added to highlight the main shape/features of the vessels as it is required by the CVA’s guidelines, which aim to remove any distortion (e.g. traces of wear) from the figures shown. This has to be done because distortions shown in unedited images can distract the readers. For optical 3D-scanners this especially concerns the inner part of the profile lines of vessels with a narrow orifice, because it can only be partially acquired as shown in Figure 6a.
Additional geometric information of other parts of the vessel, like the cross-section of the handle are estimated from the 3D-model and added to the profile line. Another benefit from using 3D-model are top-views, which can be estimated without the loss of accuracy (photographs have perspective distortion!) and as easily adopted for publication as the profile lines. A rendered top-view and its counter-part of the publication are shown in Figure 7.
The final drawing and the textual description will be available through the CVA.
As the Spectrometer can only acquire multi-spectral readings at certain spots (points) and due to time constraints (10min/reading) archaeologists selected points of interest, which are typically colors used to paint important parts of the figure, like textiles, hair, tools, etc. Figure 5.3 shows the reflectograms for “red hair” as points of interest. Therefore we can show a significant difference between the reflectrograms of different vessels between 1400 and 1500nm (infrared), which indicates, that different ingredients have been used – even though they appear the same for the human eye. An application of this analysis can help to determine fake objects in case a museum wants to add new objects to their collection. In general the acquired data and this technique in combination with a pigment database will assist in further analysis.
Summary and Outlook
Summarizing the results of our interdisciplinary, cooperative project, we could show that 3D-acquisition is an eligible tool for acquisition and documentation of ceramics from the masses of daily finds on excavations up to unique objects of museums collections. We could show that we could decrease the time and costs for documentation for such objects, especially because we have a precise, digital drawing/3D-model. It must be stressed, that even more time and costs can be saved, because the digital drawings can be easily adopted for different purposes e.g. publication in different scale and used within digital libraries. Especially the use within such libraries will enable experts to digitally analyze and compare vessels (often not on public display) of different museums. Such an analysis can be an estimation of the volume as shown in Figure 9 or a statistical analysis of the symmetry of a vessel to determine the manufacturing process (Mara 2006b). Furthermore 3D-acquisition using optical methods is radiation-free and contact-free, which can be done in-situ at excavations and in museums, which minimizes the risks to damage to objects and maximizes the accuracy.
Additionally we are currently adapting the London Charter (Beacham 2006, Web: http://www.londoncharter.org) to ensure the intellectual integrity, reliability, transparency, documentation, standards, sustainability and accessibility of the information gathered by 3D-acquisition. This has to be done, because digital data can be miss-interpreted or in worst-case miss-used, especially when it gets available on the Internet (Ogleby 2007).
Future work will be the adoption of digital libraries for textured 3D-models including registered multi-spectral readings of ceramics. Furthermore we planned to set up a database of color pigments in collaboration with experts of chemistry, similar to the work done within the project Casandra (FWF-Project P15471-MAT, Assinger 2005).
We would like to thank the Austrian Science Foundation (FWF, Project Nr. 18213), the Museum for History of Arts (Kunsthistorisches Museum Wien Web: http://www.khm.at) and the European Commission for partially supporting this publication within the CHIRON-fellowship (MEST-CT-2004-514539).
Assinger 2005, C. Asinger, P. Kammerer, E. Zolda, and P. Tatzer. Classification of Color Pigments in Hyperspectral Images. Eds: A. Hanbury and H. Bischof, Proc. of the 10th Computer Vision Winter Workshop, CVWW, pp 205-214
Beacham 2006 Beacham, R., Denard, H. and Niccolucci, F., 2006. An Introduction to the London Charter. In: M. I. et al. (ed.), The e-volution of Information Communication Technology in Cultural Heritage: where hi-tech touches the past: risks and challenges for the 21st century, Short papers from the joint event CIPA/VAST/EG/EuroMed, Archaeolingua, Budapest, Hungary.
Cosmas 2001, Cosmas, J., Itagaki, T., Green, D., Grabczewski, E., Gool, L. V., Zalesny, A., Vanrintel, D., Leberl, F., Grabner, M., Schindler, K., Karner, K., Gervautz, M., Hynst, S., Waelkens, M., Pollefeys, M., DeGeest, R., Sablatnig, R. and Kampel, M., 2001. 3D MURALE: A Multimedia System for Archaeology. In: Proceedings of the International Conference on Virtual Reality, Archaeology and Cultural Heritage, Athens, Greece, pp. 297–305Leute,1987, Archaeometry: An Introduction to Physical Methods in Archaeology and the History of Art. John Wiley & Sons.
DePiero 1996, DePiero, F. W. and Trivedi, M. M., 1996. 3-D Computer Vision Using Structured light: Design, Calibration, and Implementation Issues. Advances in Computers 43, pp. 243–278.
Gander 1994: W. Gander, G.H. Golub, and R. Strebel. Least-squares fitting of circles and ellipses. BIT, 34:558–578
Kampel 1999, Kampel, M. and Sablatnig, R., 1999. On 3d Modelling of Archaeological Sherds. In: Proceedings of the International Workshop on Synthetic-Natural Hybrid Coding and Three Dimensional Imaging, pp. 95–98.
Koch-Brinkmann 1999: Ulrike Koch-Brinkmann, Polychrome Bilder auf weißgrundigen Lekythen, 1999).
Liska 1999, Liska, C., 1999. Das Adaptive Lichtschnittverfahren zur Oberflächenkonstruktion mittels Laserlicht. Master’s thesis, Vienna University of Technology, Vienna University of Technology, Institute of Computer Aided Automation, Pattern Recognition and Image Processing Group.
Mara 2006 H. Mara, Documentation of Rotationally Symmetrical Archaeological Finds by 3D Shape Estimation ("Diplomarbeit" - Thesis), Vienna University of Technology, Inst. of Computer Aided Automation, Pattern Recognition and Image Processing Group, Tech. Report Nr: PRIP-TR-103
Ogleby 2007, Cliff Ogleby: The "Truthlikeness" of Virtual Reality Reconstructions of Architectural Heritage: Concepts and Metadata. In: Proc. Of 2nd International Workshop 3D-ARCH'2007 3D Virtual Reconstruction and Visualization of Complex Architectures, International Archives of Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol. XXXVI-5/W47, ISSN 1682-1777, Editors: F. Remondino, S. El-Hakim, 2007, ETH Zurich, Switzerland
Tosovic 2002: S. Tosovic. Adaptive 3D Modeling of Objects by combining Shape from Silhouette and Shape from Structured Light. Master’s thesis, Vienna University of Technology, Vienna University of Technology, Institute of Computer Aided Automation, Pattern Recognition and Image Processing Group, February 2002.
Wehgartner 1983, Irma Wehgartner 1983: Attisch weißgrundige Keramik, Mainz am Rhein, von Zabern 1983.
Mara, H., et al, 3D-Acquisition and Multi-Spectral Readings for Documentation of Polychrome Ceramics in the Antiquities Collection of the Kunsthistorisches Museum Vienna , in International Cultural Heritage Informatics Meeting (ICHIM07): Proceedings, J. Trant and D. Bearman (eds). Toronto: Archives & Museum Informatics. 2007. Published October 24, 2007 at http://www.archimuse.com/ichim07/papers/mara/mara.html