Petroglyph Recognition Using Self-Organizing Maps and Fuzzy Visual Language Parsing

Petroglyphs are images carved into a rock surface by prehistoric people using a symbolic or ritual language. Although they constitute an historical patrimony of inestimable value, little efforts have been devoted to the development of automated tools for their classification and interpretation. In this work we present a new algorithm for recognizing petroglyphs within scenes composed of several engraved figures. The proposal combines an unsupervised recognizer, Self-Organizing Maps (SOM), with a fuzzy visual language parser. The first classifies the petroglyph symbols extracted from a scene by using Radon transform as shape descriptor. The latter exploits the archeological knowledge about recurring patterns within scenes to solve ambiguous interpretations. The algorithm has been evaluated on a set of 50 petroglyph scenes, containing about 500 carved symbols from Mount Bego rock art site, and achieved very promising results.