2D & 3D Analysis of Inscription
Inscriptions were carved into stone whenever a culture wanted to preserve a historic incidence or achievement for future generations. This ranges from early Greek law texts in Ephesos, for example, to Buddhist stone scriptures from Shandong, and includes inscriptions on tombs or buildings bearing the names of monarchs or rulers of ancient empires from long ago. The immense cultural heritage of more than 3000 years of cuneiform writing is a source for our historic self-image, but first of all has to be collected, decoded and transcribed to be available to scientists from various disciplines.
In the case of Buddhist rock inscriptions, the scripture used to be taken from the rock by placing paper on top and rubbing the script with graphite. The result consists of white characters on a gray and rather noisy background, which makes it difficult to analyze the inscriptions: Which feature is part of a character and what is a scratch or defect caused by weathering?
After enhancing a 2D image using standard techniques, optical character recognition (OCR) still failed for most of the characters since they are not very common in modern Chinese. The scientists working on this project then implemented machine-learning methods to improve recognition of special characters known from that period.
3D Character Transcription for Ancient Greek and Cuneiform
Scientists have already collected an immense amount of 2D image data of various inscriptions, ranging from graphite rubbing to side-lit photography, all of which can be handled using almost the same methods. Nevertheless, there are better methods of obtaining data from a relief, namely a 3D scan of the surface.
The highest resolution optical 3D white light scanner, which is used in this project, not only takes the geometric information but also provides information on the color of the various objects. With all this additional data, edges can be detected without interference caused by poorly placed light sources or misleading color information. Computer scientists collaborating with mathematicians can measure the local curvature in relation to the curvature on a coarser scale and are able to separate noise from script as well as imprints of cylinder seals. Or they can simply mirror the imprints to obtain the shape of the seal. Finally, the scientists translate their calculations back into color tables and map this information onto the geometric object or, alternatively, generate a plane projection.
This kind of transcription was previously done by hand and took almost 80% of the time of an expert. Now, with the help of the computer scientists and mathematicians, assyriologists are able to read the precisely marked cuneiform as black triangles on a white background easily and can thus concentrate on the content rather than the decoding. New hypotheses can be proven on the basis of cylinder seals or the internal imprint of a clay envelope.
Jewish Cemetery Worms
Similar methods are also applied to the carving on Jewish tombstones exposed to weathering for centuries. The Jewish culture does not allow a grave to be touched after the body has been buried. For this reason, Jewish cemeteries are of great interest to geomorphologists, because, in the middle of a modern city, there lies an area untouched for centuries.
Geographers scanned the entire terrain of the "Heiliger Sand" cemetery in Worms with a midrange time-of-flight 3D laser scanner. Using georadar measurement and better deciphering of the inscriptions, this research project provides an insight into Jewish culture in the Middle Ages.
Dr. Pia Heberer - Rheinland Pfalz State Office for Care of Monuments, Mainz
Prof. Michael Brocke - Steinheim Institute, Duisburg
Prof. Olaf Bubenzer - Department of Geography, Heidelberg University