Petroglyph Classification using the Image Distortion Model

Abstract

Petroglyphs are prehistoric engravings in stone unrevealing stories of ancient life and describing a conception of the world transmitted till today. The great number of sites and the high variability in the artifacts makes their study a very complex task. Thus, the development of tools which automate the recognition of petroglyphs is essential not only for supporting archaeologist to understand petroglyph symbols and relationships, but also for the anthropologists who are interested in the evolution of human beings. However, many challenges exist in the recognition of petroglyph reliefs mainly due to their high level of distortion and variability. To address these challenges, in this paper we present an automatic image-based petroglyph recognizer that focuses on the visual appearance of the petroglyph in order to assess the similarity of petroglyph reliefs. The proposed matching algorithm is based on an image deformation model that is computationally efficient and robust to local distortions. The classification system has been applied to an image database containing 17 classes of petroglyph symbols from Mount Bego rock art site achieving a classification rate of 68 percent.