近年來，臺灣山區道路、邊坡與擋土牆受到強降雨、風化作用及地形地質構造等條件影響，常於地表及結構物留下裂縫。裂縫之分布位置、型態與聚集情形，可能反映坡體狀態、地質條件及潛在活動程度。然而，傳統裂縫調查多仰賴人工巡查與現地判釋，耗費大量人力與時間，亦難以進行大範圍且系統性的盤查。
本研究以中部山區公路沿線環景影像為基礎，建立結合深度學習與地理資訊系統之裂縫辨識及空間分析流程。首先載入既有之YOLO v12m擋土牆辨識模型，定位影像中的擋土牆範圍，再由本研究建置之RF-DETR Segmentation模型進行裂縫分割，自大量影像中擷取擋土牆裂縫資訊並建立裂縫分布資料庫，進一步與地質構造、地形條件及邊坡活動相關圖資進行疊圖分析。
研究結果顯示，本研究建立之兩階段流程可有效自大量環景影像中篩選擋土牆並萃取裂縫資訊。其中，既有YOLO v12m擋土牆辨識模型可作為大量環景影像之初步篩選工具，而本研究建置之RF-DETR Segmentation模型對於細長、斷續且背景複雜之裂縫具有較完整的分割成果。裂縫沿道路之分布並不均勻，而是集中於數個局部路段，顯示裂縫發育可能受到局部地質、地形及坡體活動條件影響。
裂縫密集區與部分潛在大規模崩塌範圍具有一定空間關聯，且部分區域位於主要斷層上盤側或褶皺構造附近。然而，部分裂縫密集區與既有圈繪之潛在大規模崩塌範圍並未完全重疊，顯示裂縫分布除可能反映已知潛在大規模崩塌外，亦可提供現有地質判釋範圍之外的輔助資訊。研究成果可作為山區裂縫快速盤查、重點區域篩選及後續坡體活動性追蹤之參考。

Mountain roads and retaining walls in Taiwan are continually affected by rainfall, weathering, geological structures, and topographic conditions. Although a single crack cannot directly indicate slope instability, the location and concentration of cracks may provide useful information about local geological conditions and potential slope activity. This study used panoramic images collected along a mountain highway in central Taiwan to establish a crack identification and spatial analysis workflow integrating deep learning and a geographic information system. Existing YOLO v12m retaining wall detection model was first used to locate retaining walls, followed by an RF-DETR Segmentation model developed in this study to segment cracks. The detected crack locations were converted into spatial data and compared with terrain, geological structures, and potential large-scale landslide areas. The results show that the workflow can efficiently screen large image datasets and identify several local crack concentration areas. Some areas corresponded with mapped geological or slope activity conditions, while others were located outside existing interpretation boundaries. The workflow may therefore support rapid crack surveys, priority-area screening, and subsequent field verification.

中華民國景觀學會（2003）。《景觀道路相關設施設計及施工參考手冊研訂》。交通部高速公路局。https://www.freeway.gov.tw/Research.aspx?RID=2

內政部國土測繪中心（2024）。國土測繪圖資服務雲：臺灣通用電子地圖。https://maps.nlsc.gov.tw/

內政部國土測繪中心（無日期）。國土測繪圖資服務雲：歷年正射影像圖層（2018–2025）〔網路地圖服務〕。2026年7月14日檢索自https://maps.nlsc.gov.tw/

江彥㚬（2026）。《應用車載全景影像與深度學習之擋土牆裂縫自動辨識與案例分析》〔未出版碩士論文〕。國立成功大學地球科學系。

何春蓀（1986）。《臺灣地質概論：臺灣地質圖說明書》。經濟部中央地質調查所。

徐明志、蔡晧川、黃筱卿、李守原、江政恩、謝孝維（2015）。擋土牆功能檢視、調查及維護。《土木水利》，42（1），73–83。https://www.ciche.org.tw/wordpress/wp-content/uploads/2018/03/DB4201-P073-%E6%93%8B%E5%9C%9F%E7%89%86%E5%8A%9F%E8%83%BD.pdf

許晉瑋（2025）。《山崩與地滑地質敏感區調查及評估》〔講習會簡報〕。2025地質敏感區與土地開發實務講習會，經濟部地質調查及礦業管理中心。

經濟部（1987）。《臺灣地體構造圖》（比例尺1：500,000）。經濟部中央地質調查所。

經濟部地質調查及礦業管理中心（2023a）。地質資料整合查詢系統。https://geomap.gsmma.gov.tw/

經濟部地質調查及礦業管理中心（2023b）。山崩地質資訊雲端服務平臺。https://landslide.geologycloud.tw/

劉桓吉、高銘健（2010）。《梨山地質圖幅及說明書：五萬分之一臺灣地質圖及說明書第19號》。經濟部中央地質調查所。

羅偉（1993）。《大禹嶺地質圖幅及說明書：五萬分之一臺灣地質圖及說明書第27號》。經濟部中央地質調查所。

羅偉、吳樂群、陳華玟（1999）。《國姓地質圖幅及說明書：五萬分之一臺灣地質圖及說明書第25號》。經濟部中央地質調查所。

羅偉、楊昭男（2002）。《霧社地質圖幅及說明書：五萬分之一臺灣地質圖及說明書第26號》。經濟部中央地質調查所。

ACI Committee 224. (2001). Control of cracking in concrete structures (ACI 224R-01). American Concrete Institute.

Anderson, T. K. (2009). Kernel density estimation and K-means clustering to profile road accident hotspots. Accident Analysis & Prevention, 41(3), 359–364. https://doi.org/10.1016/j.aap.2008.12.014

Bellahsen, N., Fiore, P., & Pollard, D. D. (2006). The role of fractures in the structural interpretation of Sheep Mountain Anticline, Wyoming. Journal of Structural Geology, 28(5), 850–867. https://doi.org/10.1016/j.jsg.2006.01.013

Berg, R. R., Christopher, B. R., & Samtani, N. C. (2009). Design and construction of mechanically stabilized earth walls and reinforced soil slopes—Volume I (FHWA-NHI-10-024). Federal Highway Administration.

Buades, A., Coll, B., & Morel, J. M. (2005). A non-local algorithm for image denoising. In 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’05) (Vol. 2, pp. 60–65). IEEE. https://doi.org/10.1109/CVPR.2005.38

Carion, N., Massa, F., Synnaeve, G., Usunier, N., Kirillov, A., & Zagoruyko, S. (2020). End-to-end object detection with transformers. In A. Vedaldi, H. Bischof, T. Brox, & J.-M. Frahm (Eds.), Computer vision—ECCV 2020 (pp. 213–229). Springer. https://doi.org/10.1007/978-3-030-58452-8_13

Dorafshan, S., Thomas, R. J., & Maguire, M. (2018). SDNET2018: An annotated image dataset for non-contact concrete crack detection using deep convolutional neural networks. Data in Brief, 21, 1664–1668. https://doi.org/10.1016/j.dib.2018.11.015

Dung, C. V., & Anh, L. D. (2019). Autonomous concrete crack detection using deep fully convolutional neural network. Automation in Construction, 99, 52–58. https://doi.org/10.1016/j.autcon.2018.11.028

Fan, C., Ding, Y., Liu, X., & Yang, K. (2025). A review of crack research in concrete structures based on data-driven and intelligent algorithms. Structures, 75, Article 108800. https://doi.org/10.1016/j.istruc.2025.108800

Li, X., Wang, W., Wu, L., Chen, S., Hu, X., Li, J., Tang, J., & Yang, J. (2020). Generalized focal loss: Learning qualified and distributed bounding boxes for dense object detection. Advances in Neural Information Processing Systems, 33, 21002–21012.

Liu, Y., Yao, J., Lu, X., Xie, R., & Li, L. (2019). DeepCrack: A deep hierarchical feature learning architecture for crack segmentation. Neurocomputing, 338, 139–153. https://doi.org/10.1016/j.neucom.2019.01.036

Mehta, P. K., & Monteiro, P. J. M. (2014). Concrete: Microstructure, properties, and materials (4th ed.). McGraw-Hill Education.

Mohammed, I. A., & Isa, N. A. M. (2025). Contrast limited adaptive local histogram equalization method for poor contrast image enhancement. IEEE Access, 13, 62600–62632. https://doi.org/10.1109/ACCESS.2025.3558506

Mohan, A., & Poobal, S. (2018). Crack detection using image processing: A critical review and analysis. Alexandria Engineering Journal, 57(2), 787–798. https://doi.org/10.1016/j.aej.2017.01.020

Nilson, A. H., Darwin, D., & Dolan, C. W. (2010). Design of concrete structures (14th ed.). McGraw-Hill.

QGIS Development Team. (2024). QGIS Geographic Information System [Computer software]. Open Source Geospatial Foundation Project. https://www.qgis.org/

Robinson, I., Robicheaux, P., Popov, M., Ramanan, D., & Peri, N. (2025). RF-DETR: Neural architecture search for real-time detection transformers [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2511.09554

Roboflow. (2024). RF-DETR Segmentation preview: Transformer-based instance segmentation with RF-DETR. https://blog.roboflow.com/rf-detr-segmentation-preview/

Sapkota, R., Cheppally, R. H., Sharda, A., & Karkee, M. (2025). RF-DETR object detection vs. YOLOv12: A study of transformer-based and CNN-based architectures for single-class and multi-class greenfruit detection in complex orchard environments under label ambiguity [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2504.13099

Sari, U. C., Sholeh, M. N., & Hermanto, I. (2020). The stability analysis study of conventional retaining walls variation design in vertical slope. Journal of Physics: Conference Series, 1444(1), Article 012053. https://doi.org/10.1088/1742-6596/1444/1/012053

Sohaib, M., Arif, M., & Kim, J. (2024). Evaluating YOLO models for efficient crack detection in concrete structures using transfer learning. Buildings, 14(12), Article 3928. https://doi.org/10.3390/buildings14123928

Tian, Y., Ye, Q., & Doermann, D. (2025). YOLOv12: Attention-centric real-time object detectors [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2502.12524

U.S. Army Corps of Engineers. (1989). Retaining and flood walls (Engineer Manual EM 1110-2-2502). U.S. Department of the Army.

Xie, Z., & Yan, J. (2008). Kernel density estimation of traffic accidents in a network space. Computers, Environment and Urban Systems, 32(5), 396–406. https://doi.org/10.1016/j.compenvurbsys.2008.05.001

Zhang, Y., & Zhao, W. (2025). Landslide identification based on characteristics of retaining wall fractures and rainfall stability analysis: A case study of fill slope in Zheng’an County, Guizhou Province. The Chinese Journal of Geological Hazard and Control, 36(2), 126–135. https://doi.org/10.16031/j.cnki.issn.1003-8035.202308027

Zuiderveld, K. (1994). Contrast limited adaptive histogram equalization. In P. Heckbert (Ed.), Graphics gems IV (pp. 474–485). Academic Press.