AUTOMATED INVESTIGATION OFARCHAEOLOGICAL VESSELS M. Kampel, H. Mara and Robert Sablatnig Vienna University of Technology, Pattern Recognition and Image Processing Group Favoritenstrasse 9/183, 1040 Vienna, Austria phone: + (43) 158801 183 64, fax: + (43) 58801 18392, email: {kampel,mara,sab}@prip.tuwien.ac.at web: www.prip.tuwien.ac.at ABSTRACT Motivated by the requirements of the present archaeology, we are developing an automated system for archaeological classification and reconstruction of ceramics. This paper shows a method to answer archaeological questions about the manufacturing process of ancient ceramics, which is impor- tant to determine the technological advancement of ancient culture. The method is based on the estimation of the pro- file lines of ceramic fragments, which can also be applied to complete vessels. With the enhancements shown in this paper, archaeologists get a tool to determine ancient manu- facturing techniques. 1. INTRODUCTION Thousands of fragments of ceramics (called sherds for short) are found at archaeological excavation sites. Ceramics are among of the most widespread archaeological finds, have been used for a short period of time for classification pur- poses. Since the 19th century, the physical characteristics of archaeological pottery have been used to assess cultural groups, population movements, inter-regional contacts, pro- duction contexts, and technical or functional constraints (ar- chaeometry). Because archaeometry of pottery still suffers from a lack of methodology, it is important to develop ana- lytical classification tools of artifacts [OTV93]. Traditional archaeological classification is based on the so-called pro- file of the object, which is the crossection of the fragment in the direction of the rotational axis of symmetry. This two- dimensional plot holds all the information needed to perform archaeological research. The correct profile and the correct axis of rotation are thus essential to reconstruct and classify archaeological ceramics. In this approach we use complete vessels, because sherds of excavations of living places have been dumped and re- used as filling material for floors and walls. Therefore sherds virtually never reassemble a complete vessels and therefore no real ground truth is known. As archaeologists are also excavating burial places, where individual unbroken ceram- ics or complete sets of sherds are found, our method can be applied on, but is not limited to, individual vessels. Conclusions about the manufacturing process can reveal important information for archaeologists, because the man- ufacturing technology gives information about development of an ancient culture, For example archaeologists determine between ceramics, that have been produced on slow or fast turning rotational plates. Another example would be an ongoing discussion between archaeologists about the existence of rotational plates for manufacturing ceramics in South America. The general opinion is that in this region the wheel was not invented and there fore ceramics were produced with a rotational plate (wheel) [WT02] on the other hand-side there is evidence that rotational plates were used [Car86]. As we use structured light as 3D-acquisition method, we can not make an assumption about the internal structure of a ceramic like [WT02], but we can estimate the surface with high resolution (0.1 mm). Therefore we can analyze the symmetry and estimate features like deviation of real surfaces with respect to a perfect symmetrical surface. Such features can help archaeologists to decide about the technological advancements of ancient cultures. The remainder of this paper is organised as follows: Sec- tion 2 briefly describes the acquisition technique used for get- ting 3D surface data of the objects. The estimation of profile lines and the way how we find the proper orientation of the vessels is described in section 3. Section 4 presents experi- mental results and discusses them. Finally, in Section 5 we draw the conclusions and comment on future work. 2. DATA ACQUISITION The acquisition method for estimating the 3D shape of a frag- ment is Shape from structured light (SfSL) [DT96], which is based on active triangulation [Bes88]. SfSL is a method which constructs a surface model of an object based on pro- jecting a sequence of well defined light patterns onto the ob- ject. The patterns can be in the form of coded light stripes [KSM02] or a ray or plane of laser light [Lis99]. In the pro- cess of calibration the parameters to describe the position of the sensors in a reference co-ordinate system and the sensor characteristics of the camera are estimated [KS01]. If the geometry between the light plane and the image is known, then each 2D image point belonging to the laser line corre- sponds to exactly one 3D point on the surface of the object [KKS96, Shi87]. This process is also called active triangu- lation [Bes88, DT96]. The volume of the fragments to be processed ranges from 3 × 3 × 3cm 3 to 30 × 30 × 50cm 3 . The Vivid 900 3D Scanner in our setup consists of the following devices: • one CCD-camera with a focal length of 14mm and a res- olution of 640 × 480 pixels, equipped with a rotary filter for color separation. • one red laser with a wavelength of 670nm and a maxi- mal power of 30mW . The laser is equipped with a gal- vanometer mirror in order to open loop control the laser beam scanning motion. Figure 1 illustrates the acquisition setup consisting of the Vivid 900 Scanner connected to a PC and the object to be recorded. Optionally the object is placed on a turntable with