臺灣山區地質活動頻繁，擋土牆若出現大規模且系統性裂縫，除可能反映牆體本身之破壞外，亦可能與道路上方坡體變形、地質構造或活動塊體等因素有關；道路鋪面裂縫則可作為輔助判斷路基及下方坡體變形之參考，因此擋土牆與道路裂縫之型態、空間分布及時序變化，可作為早期判釋地質危害之重要線索。
本研究以臺灣南部山區南橫公路玉井36 K至向陽149 K為研究範圍，採用 Ladybug5+ 車載環景影像作為主要資料來源。經影像前處理、人工篩選與標註後，建立擋土牆資料集，並以YOLOv12偵測環景影像中之擋土牆範圍；另採用RDD2022日本子資料集，訓練YOLOv12於後鏡頭影像辨識道路裂縫；再利用已訓練權重之RF-DETR於擋土牆或道路裂縫範圍內進行裂縫分割，取得像素級裂縫遮罩。YOLOv12擋土牆辨識模型之mAP@0.5為0.79，最高F1分數為0.76；後續針對部分裂縫輔以LiDAR點雲資料進行長度量測。
將裂縫位置及數量建置於GIS系統，結合地質圖資進行套疊分析，以篩選裂縫密集且具地質意義之代表路段進行案例討論，同時輔以裂縫型態與線性裂縫頻率分析，判釋裂縫分布與坡體變形及地質構造之可能關係。梅山口（Meishankou）案例位於活動塊體邊緣，擋土牆主要發育橫向長裂縫與網狀破裂，道路鋪面未見明顯裂縫，可能反映變形主要集中於擋土牆及上方坡體。甲仙（Chiahsien）案例則位於內英斷層（Neiying Fault）通過區域，裂縫集中於斷層附近，並伴隨道路裂縫、滲水等現象，顯示其可能受斷層控制與坡體變形共同影響。此外，Google歷史街景顯示，甲仙（Chiahsien）案例擋土牆裂縫早於道路裂縫形成，指示該區變形可能為漸進發展。整體而言，本研究建立結合車載環景影像、擋土牆及道路裂縫自動化辨識、GIS空間分析、時序影像比對之擋土牆裂縫分析流程，可作為判讀山區道路周邊坡體變形及地質危害早期判釋之參考。

Retaining wall cracks along mountainous roads can indicate deterioration of the wall itself and may also reflect slope deformation associated with geological structures and active landslide blocks. Conventional crack inspections rely on manual field surveys, which are time-consuming and inefficient for large-scale. This study develops an integrated workflow for retaining wall crack analysis using vehicle-mounted panoramic imagery, deep learning-based detection, GIS-based spatial analysis.
The study area covers the Southern Cross-Island Highway in southern Taiwan, extending from Yujing to kilometer post Xiangyang. Ladybug5+ panoramic images were used as the primary data source. After preprocessing, manual screening, and annotation, a retaining wall dataset was established. YOLOv12 was applied to detect retaining wall regions in panoramic images, achieving an mAP@0.5 of 0.79 and a maximum F1-score of 0.76. A second YOLOv12 model trained on the Japan subset of RDD2022 was used to detect pavement cracks in rear-view images as auxiliary information. The detected regions of interest were then segmented into binary crack masks using a pre-trained crack segmentation model, and the results were overlaid in GIS with geological and landslide-related data.
Two crack-concentrated sections were examined. The Meishankou case is located near the margin of an active landslide block and is dominated by long transverse cracks and network-like fracturing. The Chiahsien case lies where the highway crosses the Neiying Fault, with cracks accompanied by pavement cracking. Historical Google Street View images suggest that retaining wall cracks preceded pavement cracks, consistent with possible progressive deformation.

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。交通部臺灣區國道高速公路局。
朱劭謹（2026）。山區公路擋土牆裂縫分布特徵與邊坡活動關聯性之研究。國立成功大學。
交通部高速公路局（2022）。公路邊坡工程設計規範（交技〔111〕字第1115006074號）。交通部高速公路局。
何春蓀（1956）。臺南竹頭崎油田之中新世地層。臺灣省地質調查所彙刊，8，7-14。
何春蓀（1975）。臺灣地質概論：臺灣地質圖說明書。經濟部中央地質調查所。
何春蓀（1986）。臺灣地質概論：臺灣地質圖說明書（增訂第2版）。經濟部中央地質調查所。
何春蓀（1987）。中央山脈西翼及脊梁山脈區的地質。地工技術，19，81-90。
李錫堤、王源（1985）。臺灣南部橫貫公路禮觀一帶之地層及構造。地質，6（1），1-20。
宋國城、林慶偉、林偉雄與林文正（2000）。五萬分之一臺灣地質圖說明書—甲仙圖幅（圖幅51號），經濟部中央地質調查所。
經濟部中央地質調查所（2014a）。山崩與地滑地質敏感區劃定計畫書：臺南市（頁16-17）。經濟部中央地質調查所。
經濟部中央地質調查所（2014b）。山崩與地滑地質敏感區劃定計畫書：高雄市（頁18）。經濟部中央地質調查所。
經濟部中央地質調查所（2014c）。山崩與地滑地質敏感區劃定計畫書：臺東縣（頁16-18）。經濟部中央地質調查所。
內政部國土測繪中心（2026）。國土測繪圖資服務雲：WMTS 圖磚服務。檢索日期：2026年7月15日，取自https://maps.nlsc.gov.tw/。
經濟部地質調查及礦業管理中心（2026a）。地質資料整合查詢服務：WMS／WMTS 介接服務。檢索日期：2026年7月15日，取自 https://geomap.gsmma.gov.tw/gsb108-1/list_service.cfm。
經濟部地質調查及礦業管理中心（2026b）。山崩地質資訊雲端服務平臺。檢索日期：2026年7月15日，取自 https://landslide.geologycloud.tw/。
Ali, R., Chuah, J. H., Abu Talip, M. S., Mokhtar, N., & Shoaib, M. A. (2022). Structural crack detection using deep convolutional neural networks. Automation in Construction, 133, Article 103989. https://doi.org/10.1016/j.autcon.2021.103989
Angelier, J., Chu, H. T., & Lee, J. C. (1997). Shear concentration in a collision zone: kinematics of the Chihshang Fault as revealed by outcrop-scale quantification of active faulting, Longitudinal Valley, eastern Taiwan. Tectonophysics, 274(1-3), 117-143. https://doi.org/10.1016/S0040-1951(96)00301-0
Arya, D., Maeda, H., Ghosh, S. K., Toshniwal, D., & Sekimoto, Y. (2024). RDD2022: A multi-national image dataset for automatic road damage detection. Geoscience Data Journal, 11(4), 846-862. https://doi.org/10.1002/gdj3.260
Avar, B. B., & Hudyma, N. W. (2019). Earthquake surface rupture: A brief survey on interdisciplinary research and practice from geology to geotechnical engineering. Rock Mechanics and Rock Engineering, 52(12), 5259-5281. https://doi.org/10.1007/s00603-019-02006-0
Cha, Y.-J., Choi, W., & Büyüköztürk, O. (2017). Deep learning-based crack damage detection using convolutional neural networks. Computer-Aided Civil and Infrastructure Engineering, 32(5), 361-378. https://doi.org/10.1111/mice.12263
Deng, P., Yao, J., Li, C., Wang, S., Li, X., Ojha, V., & He, X. (2026). 3D modeling and automated measurement of concrete cracks via segment anything refinement and visual inertial LiDAR fusion. Computer-Aided Civil and Infrastructure Engineering, 45, Article 100019. https://doi.org/10.1016/j.cacaie.2026.100019
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
Golewski, G. L. (2023). The phenomenon of cracking in cement concretes and reinforced concrete structures: The mechanism of cracks formation, causes of their initiation, types and places of occurrence, and methods of detection—A review. Buildings, 13(3), Article 765. https://doi.org/10.3390/buildings13030765
Hsieh, Y. A., & Tsai, Y. J. (2020). Machine learning for crack detection: Review and model performance comparison. Journal of Computing in Civil Engineering, 34(5), Article 04020038. https://doi.org/10.1061/(ASCE)CP.1943-5487.0000918
Huang, S., Chen, H., Yan, L., Zou, X., Li, B., & Bi, Y. (2025). A review of the progress in machine vision-based crack detection and identification technology for asphalt pavements. Digital Transportation and Safety, 4(1), 65-79. https://doi.org/10.48130/dts-0025-0006
Jiang, J., Cao, G., Gomes, D. F., & Luo, S. (2021). Vision-guided active tactile perception for crack detection and reconstruction. In 2021 29th Mediterranean Conference on Control and Automation (MED) (pp. 930-936). IEEE. https://doi.org/10.1109/MED51440.2021.9480274
Jocher, G., Chaurasia, A., & Qiu, J. (2023). Ultralytics YOLO (Version 8.0.0) [Computer software]. GitHub. https://github.com/ultralytics/ultralytics
Kaveh, H., & Alhajj, R. (2024). Recent advances in crack detection technologies for structures: A survey of 2022-2023 literature. Frontiers in Built Environment, 10, Article 1321634. https://doi.org/10.3389/fbuil.2024.1321634
Kompanets, A., Leonetti, D., Duits, R., & Snijder, H. H. (2024). Loss function inversion for improved crack segmentation in steel bridges using a CNN framework. Automation in Construction, 170, Article 105896. https://doi.org/10.1016/j.autcon.2024.105896
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
Perda, G., Eid, M. O. I., Padkan, N., Malek, S., Morelli, L., & Remondino, F. (2025). Automated detection and mapping of pavement cracks from videos for road inspections. ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, X-1/W2-2025, 99-106. https://doi.org/10.5194/isprs-annals-X-1-W2-2025-99-2025
Priest, S. D., & Hudson, J. A. (1976). Discontinuity spacings in rock. International Journal of Rock Mechanics and Mining Sciences & Geomechanics Abstracts, 13(5), 135-148. https://doi.org/10.1016/0148-9062(76)90818-4
Robinson, I., Robicheaux, P., Popov, M., Ramanan, D., & Peri, N. (2026). RF-DETR: Neural architecture search for real-time detection transformers. International Conference on Learning Representations (ICLR 2026). https://openreview.net/forum?id=qHm5GePxTh
RIEGL Laser Measurement Systems GmbH. (2025). VMX-2HA [Brochure]. https://www.riegl.com/en-austria/products/detail/riegl-vmx-2ha
Teledyne FLIR. (2025). Ladybug5+ [Datasheet]. https://www.teledynevisionsolutions.com/categories/cameras/spherical-cameras/
Terzaghi, R. D. (1965). Sources of error in joint surveys. Géotechnique, 15(3), 287-304. https://doi.org/10.1680/geot.1965.15.3.287
Tian, Y., Ye, Q., & Doermann, D. (2025). YOLOv12: Attention-centric real-time object detectors. Advances in Neural Information Processing Systems, 38, 78433-78457. https://doi.org/10.52202/085713-2627
Yang, W., Qian, Y., Kämäräinen, J.-K., Cricri, F., & Fan, L. (2018). Object detection in equirectangular panorama. In 2018 24th International Conference on Pattern Recognition (ICPR) (pp. 2190-2195). IEEE. https://doi.org/10.1109/ICPR.2018.8546070
Yang, Y., & Mei, G. (2021). Deep transfer learning approach for identifying slope surface cracks. Applied Sciences, 11(23), Article 11193. https://doi.org/10.3390/app112311193
Yuan, Q., Shi, Y., & Li, M. (2024). A review of computer vision-based crack detection methods in civil infrastructure: Progress and challenges. Remote Sensing, 16(16), Article 2910. https://doi.org/10.3390/rs16162910
Zhang, Q., Huang, S., Wang, H., Ji, Z., Zheng, S., & Liu, Y. (2025). Segmentation detection method in tree-shaded environment for road cracks collected by inspection vehicle on WFU-Unet. Scientific Reports, 15, Article 11760. https://doi.org/10.1038/s41598-025-96219-9