'Virtual archeology' connects fragments of ancient civilization

For decades, Greek archaeologists have been trying to reproduce the picture on the wall that holds important clues about ancient Thera culture, the civilization on the island buried under volcanic ash for more than 3,500 years. before.

This extremely heavy task - more than a century away at the current speed - will soon become much easier thanks to an automated system developed by a group of computer scientists at Princeton University. in collaboration with Greek archaeologists.

According to David Dobki, Professor Phillip Y. Goldman '86 prize holder for computer science and dean at Princeton, this new technology 'has the potential to change the way people do archeology'.

Dobkin and colleagues reported their work on a paper published August 15 at Los Angeles Computer Association's SIGGRAPH annual conference, considered the leading meeting in the field of computer graphics.

'The new method really makes the computer a research partner of archaeologists,' Dobkin said, having an idea for the project after his 2006 visit to the Akrotiri archaeological site on the island. Thera, today known as Santorini.

In order to save the system, Princeton's team worked closely with archaeologists and conservationists working at Akrotiri, a very prosperous area in the Bronze Age, about 1630 years ago. AD Recreate excavated wall paintings, mosaic stations or similar archaeological objects like solving giant jigsaw puzzle, however much harder. The original object often breaks into thousands of small pieces - many of which have no color, pattern or texture to distinguish, and their edges have worn away over the centuries.

Therefore, the task of recreating the above artifacts often requires a lot of effort, because archaeologists must select the fragments and use the test to put them together. Many researchers have tried to build a computer system to automate certain parts of the work, but the effort is based on hard-to-use and money-based equipment only for trained computer experts. create.

In contrast, Princeton's system uses inexpensive equipment and is designed for archaeologists and conservators to use, not computer scientists. The system uses a combination of computer algorithms and processing systems to mimic the traditional methods of archaeologists.

Szymon Rusinkiewicz, a professor of computer science, his team played a key role in Princeton's efforts, saying: 'We imitate archaeologists' methods to the fullest extent so they can Use our system at your convenience. When fully developed, the system can reduce the time it takes to create a wall from a few years to a few months. It contributes to the liberation of archaeologists so that they can focus on other important work such as restoration, restoration and ethnographic research '.

In 2007, a team of Pricenton researchers visited Akrotiri many times, first to observe and learn from highly qualified conservationists, then to test their systems. During the three days in the island in September 2007, they handled 150 fragments with their automated system.

This is one of the wall mosaics recreated from the ancient Greek civilization Thera, buried in volcanic ash over 3,500 years ago (Photo: Princeton Graphics Group)

Although the system is still in progress, it has promising results in practice. For example, when tested with a set of fragments from the Akrotiri mural, the system found 10 of the 12 matching pairs that were known. In addition, 2 matching pairs are unknown.

'This shows that the system can work in practice,' said Tim Weyrich, a professor of postdoctoral computer science at Princeton and a principal researcher and project designer. many parts of the system. Weyrich said the team is planning another trip this fall to install a fixed system for archaeologists to use. In September he will become a professor of computer science at the University of London. The installation system used by Princeton researchers includes a flatbed scanner (commonly used to scan documents, used to scan debris surfaces), laser distance meters (using laser beams to scan the width and thickness of the debris) and a mechanized dial (allowing accurate debris rotation when inspected). These devices are connected to a laptop.

By executing a predetermined sequence of actions, a conservation expert under the guidance of an archaeologist can use the system to identify, or 'obtain' 10 fragments an hour. The first flatbed scanner is used to record high-resolution color images of debris. After that, the debris is placed on the rotating plate, and the distance meter determines the visible surface of the crown from many angles. The debris is then reversed and the above procedure is repeated.

Finally, computer software, or algorithms, that perform challenging tasks is to find the meaning of low-gain information. Princeton researchers named the software they developed as 'Griphos', in Greek meaning mystery.

An algorithm that adjusts multiple surface measurements to create a precise and complete 3-dimensional image of each fragment. Another algorithm analyzes scanned images for cracked purple or small surfaces that the distance meter ignores.

The system then incorporates all the information collected - shapes, images, and surface details - into a meticulous record of each fragment.

Weyrich said: 'The information itself is very useful for archaeologists'.

Once the fragments of an object were obtained, the system began assembling them in pairs. Using only information from burrs, the system acts as a virtual archaeologist, sorting fragments to check which pieces fit together.

Benedict Brown, his doctoral thesis, recently completed under the guidance of Risinkiewicz, is largely devoted to the mural project, saying: "The ability to quickly examine everything quickly really very helpful '.

Analyzing a pair of fragments to see if they match or not happens very quickly, in just 1 or 2 seconds. However, the time it takes to assemble a mural is quite long, because the system must consider all possible pairs of debris. In order for the system to run faster, researchers plan to incorporate some additional clues that archaeologists often use to simplify their work. That data includes information such as where the debris is found, color structure and storage status.

However, Weyrich stressed that Princeton's system will never replace the case knowledge, experience and 'soft skills' of conservation experts and archaeologists.

Weyrich said: 'Reconstructing frescoes is a very complex task, considering the status of the fragments and their number. Computers perform hard parts in this process and leave important decisions for people. '

The Princeton Graphics Group's research was funded by Kress, Seeger, Thera, Cotsen Family and National Science Organization.

Reference: "A System for Fresco Fragments : " High-Quality Volume Assignment and Matching , " authors: Benedict Brown, Corey Toler-Franklin, Diego Nehab, Michael Burns, David Dobkin, Andreas Vlachopoulos, Christos Doumas, Szymon Rusinkiewicz and Tim Weyrich. ACM Transactions on Graphics (Proc. SIGGRAPH), 27 (3) 2008.