Development of material and statistical approaches for dating castles from late antiquity to medieval Byzantine period

2023 Fiscal Year Final Research Report

• Project/Area Number: 20K20721
• Research Category: Grant-in-Aid for Challenging Research (Exploratory)
• Allocation Type: Multi-year Fund
• Review Section: Medium-sized Section 3:History, archaeology, museology, and related fields
• Research Institution: Rikkyo University
• Principal Investigator: Fukatsu Yukinori 立教大学, 文学部, 教授 (70208916)
• Co-Investigator(Kenkyū-buntansha): 小口 千明 埼玉大学, 理工学研究科, 准教授 (20312803) ; 堀田 一弘 名城大学, 理工学部, 教授 (40345426) ; 浦野 聡 立教大学, 文学部, 教授 (60211778) ; 樋口 諒 名古屋大学, 高等研究院(文), 特任助教 (70827196) ; 村上 祐子 立教大学, 人工知能科学研究科, 教授 (80435502)
• Project Period (FY): 2020-07-30 – 2024-03-31
• Keywords: リキア / ビザンツ / 城壁年代 / 弾性波速度試験機

Outline of Final Research Achievements

(1) Regarding the investigation of the stacking style of building materials, we cut out parts of the castle wall with known ages and evaluated them using 3-fold cross-validation, learned using deep learning, and optimized parameters using verification images. As a result, an identification accuracy of 95% was obtained.

(2) Regarding the degree of wear and tear on building materials due to weathering, a non-destructive investigation was conducted using the Pundit PL200 set (manufactured by FTS, including analysis software) with the aim of determining the age of the stones. As a result of the investigation, there were large differences in the numerical values depending on how the remains remained (direction, whether they were underground, etc.), and it was recognized that setting parameters was an important issue.

Free Research Field

history archeology

Academic Significance and Societal Importance of the Research Achievements

本研究は、建築学、岩石風化学、画像認識、機械学習、統計学の方法や知見を援用しつつ、 古代末期～中世ビザンツ期（3世紀～1５世紀）の城壁の建設・修築年代決定を可能にするような信頼のおける科学的方法を開発することを目的とした。その結果、特に建材の積み方様式調査については、データの蓄積によって科学的評価を得る可能性があるという結論に達した。
年代が明らかな城壁の部分を切り出して3分割交差検証で評価、ディープラーニングで学習し検証用画像でパラメータを最適化した。その結果、95%の識別精度が得られた。