We propose a method for analyzing 3D-scanned Lithic artifacts and creating a rich and accurate description. When studying and documenting Lithic artifacts, archaeologists typically measure several basic metrics, such as the width, length, width/length ratio and the location of the maximum width. In addition several studies include manual extraction of the number of scars and their angle. In order to provide a richer description, few selected artifacts are usually hand drawn by a trained artist. These hand drawings are affected by the artist's subjective impression and drawing technique and in addition, are time-consuming and costly. Recent work involved scanning of artifacts in 3D and automatically calculating the basic metric attributes in an accurate and objective way. Our method combines the accuracy and objectivity of the metric methods with the richness of the descriptive methods.The proposed method is based on detection of the ridge lines and segmentation of the scars. We first calculate the maximum curvature value of each vertex on the 3D mesh. We then post-process the curvature function in order to detect the ridge lines, which are defined as the local-maxima of the curvature value in the direction of the max-curvature. We then use the detected ridge lines to segment the scars, by first performing geodesic clustering of the surface of the mesh and then optimizing the resulting segmentation using a graph-cut process.The goal of our method is twofold. First, the detected ridge lines can be used as an alternative to the traditional hand-drawings. Our method generates 2D renderings of the main views of the artifact, with the detected ridge lines clearly marked. The second goal of our method is to extract rich and accurate descriptors (or "features") that can be used in research involving automatic clustering and classification of artifacts. Because of the knapping technology that was used in making the lithic tools, in which flakes were removed from the flint core, creating well-defined scars and ridges (intersection of two scars) on the surface of the object, we believe that features that describe these scars and ridges, will enable good clustering and classification performance.We demonstrate our method on various types of Lithic artifacts from Middle Paleolithic sites in the Southern Levant. We evaluate the accuracy of the ridge detection and scars segmentation, show the resulting 2D renderings and compare them to hand-drawings (when available).
