ROBUST 3D RECONSTRUCTION OF ARCHAEOLOGICAL POTTERY BASED ON CONCENTRIC CIRCULAR RILLS Martin Kampel, Robert Sablatnig and Hubert Mara Pattern Recognition and Image Processing Group Institute for Automation, Vienna University of Technology, Favoritenstr. 9/183/2, A-1040 Vienna, Austria Fax: +43 (1) 58801 183 92; email: {kampel,sab,mara}@prip.tuwien.ac.at ABSTRACT In the area of archaeology surveying, documentation, classification, archivation and reconstruction of pottery, which is often found as thousands of fragments - so called sherds - at excavations is of major concern since statistics about social, cultural and technological status of a population can be made out of this information. Therefore we are developing an acquisition, classification and reconstruction system that is inspired by the archaeological methods. As a result of the manual manufacturing process, potter’s wheel made archaeological objects contain concentric rills on their surface. Together with the axis of rotation, which is the cross-section of the fragment in the direction of the rotational axis of symmetry, these rills form the basis for a robust 3d reconstruction. We present a fully automated approach to pottery reconstruction based on the fragment’s rills. We demonstrate the method and give results on synthetic and real data. 1. INTRODUCTION Ceramics are one of the most widespread archaeo- logical finds and are a short-lived material. This property helps researchers to document changes of style and ornaments. Especially ceramic vessels, where shape and decoration are exposed to constantly changing fashion, not only allow a basis for dating the archaeological strata, but also provide evidence of local production and trade relations of a community as well as the consumer behavior of the local population [OTV93]. The traditional method of documentation of fragments is a drawing of the profile line, which is an intersection of the fragment along the axis of symmetry (also called rotational axis), which can be found for fragments manufactured on rotational plates [YM97]. Finding this axis of rotation and drawing the profile line by hand requires expert knowledge and a certain amount of time. Therefore we are developing a fully automated system for acquisition and documentation of profile lines using a 3D scanner based on structured light [KS03,MSKS04]. The range- and pictorial information of a pottery fragment recorded by the 3D scanner serves as the basis for the classification and reconstruction process. The most important step is the determination of the profile of the fragment in the so-called orientation step. The term orientation describes the exact positioning of the fragment on the original vessel with the help of the axis of rotation. To automate this process, the profile has to be determined in the same way as in the manual documentation. The profile is rotated by the original axis of rotation, thus measurements like volume can be estimated. Thus the process of documenting a fragment is improved since the important steps measuring, drawing, and describing are automated. With the help of 3-D data, the profile of the fragment is constructed. The frontal view is represented with the help of the pictorial information of the surface of the fragment and the surface model. This representation can be used for publication or for retrieval from the database, put on the Internet by other users. This will enable it to publish both the profile of the fragment and a virtual reconstruction of the whole vessel. The paper is organized as follows: In Section 2 details on data acquisition and the rill-based orientation tech- nique are given, Section 3 comments on the automatic profile generation are given and Section 4 shows results and finally a conclusion is given. 2. DATA PROCESSING The acquisition method for estimating the 3D-shape of a fragment is shape from structured light [PT96], which is based on active triangulation [Besl88]. We used the Vivid 900 3D Scanner developed by MINOLTA. Op- tionally the object is placed on a turntable with a diameter of 40cm, whose desired position can be specified with an accuracy of 0.1°. The 3D Scanner works on the principle of laser triangulation combined with a color CCD image. It is based on a laser-stripe but a galvanometer mirror is used to scan the line over the object.