考古陶器紋飾辨識－以Lapita文化為例

Identifying Lapita Motifs Based on Pattern Recognition Technology

鄭雯方(Man-Fong Cheng)

指導教授 ： 丁肇隆

共同指導教授 ： 張瑞益

國立臺灣大學/工學院/工程科學及海洋工程學研究所/碩士(2015年)

https://doi.org/10.6342/NTU.2015.02815

摘要

科技考古為透過資訊科技輔助考古的領域，在近年來此領域日趨重視，也吸引學者們紛紛投入。過去在考古陶片的研究上，多以碎片拼接為主，然而考古陶片的紋理具有文化意義，為考古分析工作之重點，以資訊技術協助紋飾分析的流程，除可提升元數據（metadata）分析的速度，亦能在考古碎片拼接中，加入紋路特徵提升準確度。因此本研究基於南太平洋Lapita文化之考古紋飾，配合其紋飾分析規範，開發一套裝飾單元（unit）辨識系統。
系統流程分為三個部分，首先進行影像前處理與裝飾單元分割，接著以形狀上下文（shape context）與線條數統計量來表示圖像特徵，最後計算裝飾單元類別相似度，並給予使用者推薦類別之排序。實驗測試樣本包含了24類裝飾單元，分別以260個測試樣本進行分類，其辨識率高達99.6%，而平均每個單元處理時間為47.48毫秒。此外，當加入本研究之單元辨識結果當作陶片拼接特徵，可成功提升拼接色彩已消逝的Lapita考古碎片之效能。在本研究中，開發了首個 Lapita紋路單元辨識系統，其可有效減少考古工作流程的人力成本，並提升效率與效能。

關鍵字

科技考古 ； 考古紋飾辨識 ； Lapita 裝飾單元分類 ； Lapita文化 ； 考古碎片拼接

並列摘要

Recently, researchers have shown an increasing interest in archaeological science, which strives for using information technology to assist analysis of historical remains. Previous research in archaeological pottery has concentrated on shape features of pottery shreds. However, there has been little discussion about the patterns on this topic. Patterns contain cultural meaning and uniqueness, which can both improve the efficiency of archaeological metadata analysis and the accuracy of shreds assembly. In this thesis, we have developed a system of unit recognition for pattern from the Lapita pottery, which is based on the standard of Lapita motif coding.
There are three steps in the process of our system. First stages comprise image preprocessing and unit segmentation. Then, shape context and histogram of stroke count features are then conducted to extract features from units. Eventually, units were ranked in a recommended sequence for user selection. The evaluation of unit recognition was assessed from the classification of 260-sampled units into 24 categories and the accuracy is 99.6% for the top-1 recommendation. And the average process time for each unit is 47.48 milliseconds. In addition, the effectiveness of reconstructing Lapita fragments was enhanced while adding our unit recognition results to reconstruct color-faded pottery fragments. In conclusion, the first Lapita unit recognition system is developed in this thesis, which facilitates the archaeological process by its high efficiency and effectiveness.

並列關鍵字

Archaeological Science ； Archaeological Pattern Recognition ； Lapita Unit Classification ； Lapita Culture ； Archaeological Fragment Reconstruction

參考文獻

[29] L. S. Yu, H. J. Ku, C. L. Ting, R. I. Chang, Y. C. Wang, and J. M. Ho, "Content-Based Feature Matching for Fragment Reassembly of Ceramic Reconstruction," in Web Intelligence （WI） and Intelligent Agent Technologies （IAT）, 2014 IEEE/WIC/ACM International Joint Conferences on, 2014, pp. 344-351.

[4] C. A. Bollong, "Analysis of site stratigraphy and formation processes using patterns of pottery sherd dispersion," Journal of Field Archaeology, vol. 21, pp. 15-28, 1994.

[7] L. Zhu, Z. Zhou, and D. Hu, "Globally consistent reconstruction of ripped-up documents," Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 30, pp. 1-13, 2008.

[8] M. Werman and D. Weinshall, "Similarity and affine invariant distances between 2d point sets," Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 17, pp. 810-814, 1995.

[11] E. Justino, L. S. Oliveira, and C. Freitas, "Reconstructing shredded documents through feature matching," Forensic science international, vol. 160, pp. 140-147, 2006.