隧道襯砌影像判釋與異狀特徵化技術暨 安檢資料庫之開發

Interpreting and Characterizing Cracks from Tunnel Lining Image and Associated Development of Structural Safety Inspection Database

林信宏(Sin-Hong Lin)

指導教授 ： 王泰典

國立臺北科技大學/工程學院/資源工程研究所/碩士(2010年)

https://doi.org/10.6841/NTUT.2010.00379

摘要

襯砌表面影像為現代化隧道結構安全檢測作業關鍵作業項目，為判釋與記錄襯砌表面異狀分佈的基礎，更為提供襯砌結構安全評估的基本資訊。在檢測作業中常使用雷射掃描儀器取得襯砌表面影像，再利用人眼視覺方式對影像中異狀繪製成異狀展開圖，繪製過程不僅耗日費時，常需有經驗之工程師判釋影像中的異狀以及相關的肇因診斷作業。然而前、後期取得的襯砌表面影像攝錄條件不同，異狀的判釋與繪製，常需耗費大量的人力與時間，且因整體影像亮度不均勻與解析度清晰的程度不一，獲得的異狀展開圖的基準不盡相同，加大了比較異狀變異、研判異狀肇因的困難度。且因異狀展開圖為一圖形化資料，需仰賴有經驗之工程師判讀與解釋，面對現代化的維護管理作業，維護管理工作量急速上升情況下，需要更簡潔且有效的描述異狀展開圖內所隱含的資訊。
因此，本研究發展一監督式學習影像處理技術，透過異狀處與背景區域明暗灰階值的差異，偵測襯砌表面影像中的裂縫分佈，並依據前、後期取得的不同襯砌影像或是同一影像中不同區域的明暗與清晰特性，動態更新判定異狀所需的灰階門檻值，更有效地偵測異狀的分佈。繼而應用Hough轉換，依據異狀分佈特性將其映射至極座標獨立向量參數空間，可取得異狀空間分佈的數字化資訊。文中並透過數個已知肇因的異狀分佈試驗及案例，獲得不同的數字化資訊。另本研究建立隧道安檢資料庫之雛形，收集數個隧道基本資訊，整合影像及異狀數字化資訊，討論目前資料常見及後續可能之資料處理呈現方式，此資料庫可謂能提供後續建立異狀診斷以至於安全評估專家系統的有利基礎。

關鍵字

隧道襯砌 ； 異狀型態 ； 影像判釋 ； 特徵化 ； 數字化 ； 資料庫

並列摘要

Due to need to record abundant lining surface data, the layout image of a lining surface is essential to structural safety inspections of operating tunnels, and subsequent evaluation of tunnel global stability of the lining structure and local spalling of lining materials. Diverse anomalies appeared in the lining surface, i.e., the tunnel integrity was poorer than usual, which resulted in distinct outcomes, and, even worse, erroneous interpretations of anomalies. Experienced engineers always need to identify anomalies from layout images of the lining surface, and then produce an anomaly layout of the tunnel lining. The anomaly layout of a tunnel lining shows the spatial distribution of anomalies, such as cracks, leakage, and efflorescence, and is very useful reference for tunnel safety evaluations and associated diagnosis of the causes of lining anomalies. However, crack detection is time-consuming and diagnosing causes of lining anomalies is complex. These two factors must be improved to enhance the performance of tunnel inspection technologies and tunnel maintenance. To effectively indentify information from a layout image, this study employs a supervised-learning image-interpretation technology to first develop an anomaly detection method and then uses the Hough transform to characterize crack distribution patterns.

並列關鍵字

Tunnel lining ； Tunnel anomaly ； Image Detection ； Characterizing ； Database

參考文獻

[28] 王泰典，「探討襯砌異狀之類別與診斷」，營建知訊，304，2008，22-31。

[1] A. Haack, J. Schreyer and G. Jackel, "State-of-the-art of non-destructive testing methods for determining the state of a tunnel lining," Tunnelling and Underground Space Technology, vol. 10, no. 4, pp. 413-431.

[2] A. Ito, Y. Aoki and S. Hashimoto, “Accurate Extraction and Measurement of Fine Cracks from Concrete Block Surface Image,” IECON Proceedings(Industrial Electronics Conference), 2002, pp. 2202-2207.

[4] D. L. Swets and J. Weng, “Discriminant eigenfeatures for image retrieval” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 18, No. 8, pp.831-836.

[5] E. F. Codd, “A Relational Model of Data for Large Shared Data Banks,” Communications of the ACM, Vol. 13, No. 6, 1970, pp. 377-387.