<![CDATA[In archaeology, typological research methods have long been used as a reliable methodology to estimate the relative ages of artifacts and clarify their genealogical relationships. There is, however, a disadvantage to typological research methods—the researcher’s subjectivity cannot be eliminated during the analysis process. This study aimed to provide an objective typological index by applying data science to typological research. Techniques known as “feature extraction” and “unsupervised learning” were used to recognize the patterns and visualize the data. Thereby, the study is expected to help clarify the laws hidden in the iconographic data of tiles. An experiment was performed to analyze the patterns on the eaves tiles of ancient Japanese roof tiles (from Fujiwara and Heijo Palace), which are the authors’ specialty. Results revealed that matching local features focusing on edges was effective in detecting similarities between tiles and extracting differences in the general framework of the pattern structure. Furthermore, the multidimensional scaling method and phylogenetic tree were utilized to estimate the age and place of origin of each tile, which is a crucial task in archaeology. In general, the results obtained were in accordance with those of previous studies]]>
Copyrights © 2024
