簡易檢索 / 詳目顯示

回結果列表
研究生： 陳律志
Chen, Lu-Chih
論文名稱： 以點雲重建橋梁構件三維外框模型之自動化框架
Automated Structural Modeling of Bridge Components from Point Clouds
指導教授： 饒見有
Rau, Jiann-Yeou
學位類別： 碩士
Master
系所名稱： 工學院 - 測量及空間資訊學系
Department of Geomatics
論文出版年： 2026
畢業學年度： 114
語文別： 中文
論文頁數： 84
中文關鍵詞： 橋梁點雲 、自動化建模 、點雲配準 、Random Forest 、Point2CAD
外文關鍵詞： Bridge Point Clouds, Automated Modeling, Point Cloud Registration, Random Forest, Point2CAD
相關次數： 點閱：80  下載：0 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 近年來橋梁檢測逐漸朝向數位化與智慧化發展，為提升橋梁維護效率及建立完整之檢測資料，交通部運輸研究所近年推動橋梁三維影像模型作業程序，透過人工建立橋梁CAD 模型，並將模型劃分為固定區塊以貼附正射影像，供檢測人員記錄裂縫、剝落及鋼筋外露等病害位置。然而此流程仍需投入大量人工進行橋梁三維模型建置，不僅耗時費力，也限制了橋梁數位化檢測的效率。因此，如何由橋梁點雲自動建立具有結構語意之三維模型，已成為智慧橋梁檢測與數位基礎設施管理的重要研究課題。
    因此本研究提出一套橋梁點雲結構建模之自動化框架，針對橋梁點雲進行構件分割、結構辨識以及CAD 模型建立。首先利用橋梁構件之幾何、方向、位置與顏色等等特徵以建立橋梁點雲特徵描述子，並透過Random Forest 模型將橋梁點雲自動分割為多種橋梁構件。而為提升不同橋梁案例間之適應能力，本研究亦針對橋梁主方向與結構空間關係進行分析，以強化模型對不同橋型之辨識能力。
    完成橋梁構件分割後，本研究基於Point2CAD 概念針對不同橋梁構件預先建立對應之CAD 模板，並依據橋梁構件之空間位置與結構特性，選擇相對應之CAD 模型進行配準與擬合，接著透過粗配準、ICP 配準以及幾何擬合方法逐步提升CAD 模板與原始點雲間之對應精度，使模型能更符合實際橋梁結構。相較於傳統人工建模流程本研究所提出之方法可有效降低建模過程中之人工干預程度，並提升建模效率與一致性。
    實驗結果顯示，本研究所提出之方法能有效完成橋梁點雲之自動化結構建模，並於多組橋梁案例中獲得穩定之構件分割與模型重建成果。相較於傳統人工建模流程，本研究可有效降低人工作業成本，提升建模效率與一致性，未來可應用於智慧基礎設施管理、橋梁數位分身並延伸CAD 自動化建模等相關領域。

    With the increasing demand for bridge inspection and maintenance, three-dimensional (3D) bridge models have become increasingly important for digital infrastructure management. Meanwhile, rapid advances in photogrammetry have made bridge point cloud acquisition more accessible. However, automatically reconstructing semantic 3D models from large-scale bridge point clouds remains challenging, and traditional bridge modeling still relies heavily on manual operations that are time-consuming and experience-dependent.
    Therefore, this study proposes an automated framework for structural modeling of bridge components from point clouds. Geometric, directional, positional, and color features are extracted to construct feature descriptors, and a Random Forest model is employed to segment bridge point clouds into different components. In addition, principal bridge direction and structural spatial relationships are analyzed to improve adaptability across different bridge types.
    After component segmentation, CAD templates are assigned according to component type based on the Point2CAD concept. The selected templates are then refined through coarse registration, ICP registration, and geometric fitting to improve alignment with the original point clouds. The proposed framework effectively reduces manual intervention while improving modeling efficiency and consistency.
    Experimental results demonstrate that the proposed framework achieves reliable bridge component segmentation and CAD model reconstruction across multiple bridge datasets. Compared with conventional manual modeling, the proposed method significantly reduces labor requirements while improving modeling efficiency and consistency. The framework also shows potential for applications in digital twins and smart infrastructure management.

    摘要 i Abstract ii 英文延伸摘要 iii 誌謝 vi Table of Contents vii List of Tables ix List of Figures x Chapter 1 緒論. 1 1.1. 論文介紹 1 1.2. 研究動機 1 1.3. 研究架構 2 Chapter 2 文獻回顧. 4 2.1. 橋梁點雲建模 4 2.2. 點雲前處理與特徵描述 5 2.2.1. 點雲前處理 5 2.2.2. 局部幾何特徵 6 2.2.3. 位置、方向與顏色特徵 6 2.2.4. Random Forest 分類與特徵使用 7 2.3. 點雲至 CAD 模型轉換 7 2.4. 點雲擬合 8 Chapter 3 研究資料與方法. 9 3.1. 橋梁資料集 9 3.1.1. 影像與相機資訊 9 3.1.2. 控制點與座標系統 10 3.1.3. 橋梁點雲資料 11 3.2. 點雲前處理 11 3.2.1. 點雲降採樣 12 3.2.2. 雜點移除 13 3.2.2.1. 統計式離群點移除（SOR）13 3.2.2.2. 鄰近點一致性檢查（KNN）14 3.3. Random Forest Segmentation 15 3.3.1. 訓練資料集 16 3.3.2. 橋梁構件類別定義 17 3.3.3. 特徵計算與特徵可分性分析 21 3.3.3.1. RF4PCC-based Feature Extraction 21 3.3.3.2. Additional Feature Extraction 24 3.3.4. Random Forest 26 3.3.5. 橋梁構件分離（DBSCAN）28 3.4. 梁構件分離 29 3.5. Point2CAD Transformation 30 3.5.1. RANSAC-based Plane Extraction 31 3.5.2. 點雲補面 32 3.5.3. Point2CAD 原理 33 3.5.4. 輔助平面移除 35 3.6. Point Cloud-CAD Fitting 36 3.6.1. 資料匯入 37 3.6.2. 粗對準 37 3.6.2.1. 質心平移 37 3.6.2.2. 基於 BBOX 之主軸尺度縮放 38 3.6.2.3. PCA 主軸方向對齊 39 3.6.3. 初步 ICP 剛體配準 39 3.6.4. 細對準 40 3.6.4.1. 基於對應點之主軸尺度縮放 40 3.6.4.2. ICP2 細部配準 41 3.6.4.3. 迭代收斂判斷 41 3.6.5. CAD 幾何轉換與模型輸出 41 3.7. 精度評估指標 42 3.7.1. Random Forest 構件分類精度 42 3.7.2. 主梁與橫隔梁分離精度 43 3.7.3. Point Cloud–CAD Fitting 擬合精度 44 Chapter 4 實驗成果與分析. 46 4.1. Random Forest Segmentation 46 4.1.1. 模型精度 46 4.1.2. 構件分類成果 47 4.2. 梁構件分離 49 4.2.1. 主梁與橫隔梁分類成果 49 4.3. Point2CAD Transformation 51 4.4. Point Cloud–CAD Fitting 54 4.4.1. 擬合流程成果 54 4.4.2. 擬合精度評估 59 4.4.3. 擬合失敗案例分析 63 Chapter 5 結論與未來建議. 66 5.1. 研究貢獻 66 5.2. 未來建議 67 參考文獻 69

    1. M. Rashidi, M. Mohammadi, S. S. Kivi, M. M. Abdolvand, L. Truong-Hong, and B. Samali, “A decade of modern bridge monitoring using terrestrial laser scanning: Review and future directions,” Remote Sensing, vol. 12, no. 22, p. 3796, 2020.

    2. M. Mehranfar, H. Arefi, and F. Alidoost, “A projection-based reconstruction algorithm for 3d modeling of bridge structures from drone-based point cloud,” in The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. XLVI-4/W1-2021, pp. 77–83, 2021.

    3. J. H. Lee, J. J. Park, and H. Yoon, “Automatic bridge design parameter extraction for scan-to-bim,” Applied Sciences, vol. 10, no. 20, p. 7346, 2020.

    4. Z. Jiang, X. Shen, M. H. Ibrahimkhil, K. Barati, and J. Linke, “Scan-vs-bim for real-time progress monitoring of bridge construction project,” in ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. X-4/W3-2022, pp. 97–104, 2022.

    5. R. Zhang, J. Wang, X. Li, and X. Chen, “Deep-learning-based point cloud semantic segmentation: A survey,” Electronics, vol. 12, no. 17, p. 3642, 2023.

    6. M. Weinmann, B. Jutzi, S. Hinz, and C. Mallet, “Semantic point cloud interpretation based on optimal neighborhoods, relevant features and efficient classifiers,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 105, pp. 286–304, 2015.

    7. J. Demantké, C. Mallet, N. David, and B. Vallet, “Dimensionality based scale selection in 3d lidar point clouds,” in The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. XXXVIII-5/W12, pp. 97–102, 2011.

    8. T. Hackel, J. D. Wegner, and K. Schindler, “Contour detection in unstructured 3d point clouds,” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1610–1618, 2016.

    9. L. Breiman, “Random forests,” Machine Learning, vol. 45, no. 1, pp. 5–32, 2001.

    10. N. Abreu, K. Khoshelham, H. Tran, and A. Kealy, “Procedural point cloud modelling in scan-to-bim and scan-vs-bim applications: A review,” ISPRS International Journal of Geo-Information, vol. 12, no. 7, p. 260, 2023.

    11. Y. Liu, A. Obukhov, J. D. Wegner, and K. Schindler, “Point2cad: Reverse engineering cad models from 3d point clouds,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 3763–3772, 2024.

    12. F. Bosché, “Automated recognition of 3d cad model objects in laser scans and calculation of as-built dimensions for dimensional compliance control in construction,” Advanced Engineering Informatics, vol. 24, no. 1, pp. 107–118, 2010.

    13. P. J. Besl and N. D. McKay, “A method for registration of 3-d shapes,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 14, no. 2, pp. 239–256, 1992.

    14. X. Wang et al., “Dynamic downsampling algorithm for 3d point cloud map based on improved voxel filtering approach,” Applied Sciences, vol. 14, no. 8, p. 3160, 2024.

    15. R. B. Rusu, Z. C. Marton, N. Blodow, M. Dolha, and M. Beetz, “Towards 3d point cloud based object maps for household environments,” in Robotics and Autonomous Systems, 2008.

    16. T. Cover and P. Hart, “Nearest neighbor pattern classification,” IEEE Transactions on Information Theory, vol. 13, no. 1, pp. 21–27, 1967.

    17. T. Hackel, J. D. Wegner, and K. Schindler, “Towards 3d point cloud classification: A new benchmark and feature-based method,” ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 2016.

    18. M. Ester, H.-P. Kriegel, J. Sander, and X. Xu, “A density-based algorithm for discovering clusters in large spatial databases with noise,” Proceedings of the Second International Conference on Knowledge Discovery and Data Mining, pp. 226–231, 1996.

    19. M. A. Uy, M. Dahnert, M. Tatarchenko, and L. Guibas, “Point2cad: Reverse engineering cad models from 3d point clouds,” arXiv preprint arXiv:2103.12288, 2021.

    20. M. A. Fischler and R. C. Bolles, “Random sample consensus: A paradigm for model fitting with applications to image analysis and automated cartography,” Communications of the ACM, vol. 24, no. 6, pp. 381–395, 1981.

    21. scikit-learn developers, “Precision, recall, f-measure and support.” https://scikit-learn.org/stable/modules/generated/sklearn.metrics.precision_recall_fscore_support.html, 2026. Accessed: 2026-06-01.

    22. O. Rainio, J. Teuho, and R. Klén, “Evaluation metrics and statistical tests for machine learning,” Scientific Reports, vol. 14, 2024.

    23. CloudCompare, “Cloud-to-mesh distance.” https://www.cloudcompare.org/doc/wiki/index.php/Cloud-to-Mesh_Distance. Accessed: 2026-06-01.

    24. CloudCompare, “Cloud-to-cloud distance.” https://www.cloudcompare.org/doc/wiki/index.php/Cloud-to-Cloud_Distance. Accessed: 2026-06-01.

    QR CODE