本研究旨在發展一套結合深度學習與攝影測量技術之鋪面破壞量化方法，針對常見的補綻、龜裂與縱橫向裂縫三類鋪面破壞進行像素級語意分割與世界坐標下之幾何量測。為強化模型於實務場景之應用能力，本研究採用車載相機進行連續影像蒐集，透過建立三個SwinB-UperNet語意分割模型分別建立補綻、龜裂以及縱橫向裂縫二元遮罩，並結合運動推斷結構（Structure from Motion, SfM）估算相機外方位參數與建立近似數值高程模型（Digital Elevation Model, DEM），將鋪面破壞像素轉換至世界坐標系。進一步以幾何向量化方式計算補綻與龜裂面積，縱橫向裂縫則透過骨架化與像素排序後計算其主軸長度。
實驗結果評估三個語意分割模型在測試資料上的精度，補綻模型達F1-score 0.852，龜裂與縱橫向裂縫則分別為0.702與0.635。實驗結果亦顯示使用SfM能夠改正相機外方位參數並降低投影誤差。另外，藉由近似DEM方法能夠降低流程複雜度，直接從相機外方位近似得到DEM，取代點雲操作。為進一步驗證整體流程之效能，本研究選取四段總長超過1500公尺之實測路段進行評估，將模型預測與人工標註所得之鋪面破壞像素分別投影至世界坐標系後進行比較。結果顯示，在補綻方面，Precision與Recall在三個路段皆高於0.9，顯示模型可穩定量測連續補綻之面積規模，惟陰影遮蔽下仍易出現漏判。縱橫向裂縫長度方面則展現良好召回能力，各路段Recall皆超過0.8，能有效量測路段中縱橫向裂縫的位置與長度；但部分補綻邊界或線狀污漬可能誤判為縱橫向裂縫，影響Precision表現。龜裂則因邊界模糊與形態複雜，於密集縱橫向裂縫與遠距離觀測情境下常出現誤授與漏授，導致部分路段誤差比例達50%以上。
綜合而言，本研究所建構之處理流程可將語意分割成果轉換為幾何量化結果，並可於連續影像條件下評估鋪面破壞之空間分布與規模，有助於提升鋪面破壞調查之路段覆蓋率與客觀性，為鋪面管理系統提供量化評估與維護決策之參考依據。

This study develops an integrated framework that combines deep learning and photogrammetric techniques to quantify pavement distress. Three common types of pavement damage—patching, alligator cracking, and longitudinal/transverse cracking—are addressed through pixel-level semantic segmentation and geometric measurement in world coordinates. To ensure applicability in real-world conditions, continuous pavement images were collected using a vehicle-mounted camera. Three SwinB-UperNet semantic segmentation models were trained to generate binary masks for patching, alligator cracking, and longitudinal/transverse cracking. These masks were then projected into world coordinates by estimating camera exterior orientation parameters through Structure from Motion (SfM) and constructing an approximate Digital Elevation Model (DEM). Patching and alligator cracking areas were quantified via vectorization, while longitudinal/transverse crack lengths were calculated using skeletonization and pixel ordering.
Experimental evaluation demonstrated F1-scores of 0.852 for patching, 0.702 for alligator cracking, and 0.635 for longitudinal/transverse cracks. Results further indicated that SfM effectively refined exterior orientation parameters, thereby reducing projection errors, while the approximate DEM method simplified the workflow by avoiding dense point cloud processing. To validate the proposed pipeline, four road segments totaling more than 1.5 km were tested by projecting both model predictions and manual annotations into world coordinates for comparison. For patching, precision and recall consistently exceeded 0.9 across most road segments, indicating stable performance in measuring patching areas, although shadow occlusion occasionally caused omission errors. For longitudinal/transverse cracking, recall was consistently above 0.8, showing strong capability in detecting crack locations and lengths, though misclassification of patch boundaries and surface stains affected precision. Alligator cracking exhibited higher variability due to ambiguous boundaries and complex morphology, with error ratios exceeding 50% in some road sections.
Overall, the proposed framework successfully transforms semantic segmentation results into geometric measurements, enabling the evaluation of spatial distribution and extent of pavement distress under continuous imagery. This approach enhances the coverage and objectivity of pavement surveys and provides quantitative indicators to support pavement management systems and maintenance decision-making.

交通部，2019。《公路養護手冊》。臺北市：交通部公路總局。
曾志煌、陳茂南、翁國和、林志棟、康仕仲、蔣子平、周建成、張家瑞、陳建達、紀宏霖、楊政錤、林子傑、李麗芳，2010。路面缺陷影像辨識技術之研發。台北：交通部運輸研究所。（MOTC-IOT-98-EDB004；ISBN 978-986-02-4289-8）
黃敏彥 、何旻哲、林志棟、許世旻，2017。智慧車導入半自動 PCI 檢測，台灣公路工程期刊，第四十三卷第三期，2-26。
ASTM International, 2016. Standard Practice for Roads and Parking Lots Pavement Condition Index Surveys, ASTM D6433-16. West Conshohocken, PA: ASTM International.
Bajaj, R., Al-Sabbag, Z. A., Yeum, C. M., & Narasimhan, S. (2024). 3D Dense Reconstruction for Structural Defect Quantification. ASCE OPEN: Multidisciplinary Journal of Civil Engineering, 2(1), 04024001.
Bianco, S., Ciocca, G., & Marelli, D. （2018）. Evaluating the performance of structure from motion pipelines. Journal of Imaging, 4（8）, 98.
Bradski, G. The OpenCV Library. Dr. Dobb's Journal of Software Tools, 2000
Coenen, T. B., & Golroo, A. (2017). A review on automated pavement distress detection methods. Cogent Engineering, 4(1), 1374822.
Cui, L., Qi, Z., Chen, Z., Meng, F., & Shi, Y. （2015, August）. Pavement distress detection using random decision forests. In International Conference on Data Science （pp. 95-102）. Cham: Springer International Publishing.
Danyo, A., Dontoh, A., & Aboah, A. (2025). An improved ResNet50 model for predicting pavement condition index (PCI) directly from pavement images. Road Materials and Pavement Design, 1-18.
De Blasiis, M. R., Di Benedetto, A., Fiani, M., & Garozzo, M. (2020). Assessing of the road pavement roughness by means of LiDAR technology. Coatings, 11(1), 17.
Duane, C. B. (1971). Close-range camera calibration. Photogramm. Eng, 37(8), 855-866.
Du, Z., Yuan, J., Xiao, F., & Hettiarachchi, C. (2021). Application of image technology on pavement distress detection: A review. Measurement, 184, 109900.
Guan, J., Yang, X., Ding, L., Cheng, X., Lee, V. C., & Jin, C. (2021). Automated pixel-level pavement distress detection based on stereo vision and deep learning. Automation in Construction, 129, 103788.
Guo, R., Yu, Z., & Zhou, Y. (2020). Development and preliminary evaluation of a varying-speed road profiler. Journal of Testing and Evaluation, 48(5), 3479-3489.
Hadjidemetriou, G. M., Christodoulou, S. E., & Vela, P. A. （2016, April）. Automated detection of pavement patches utilizing support vector machine classification. In 2016 18th Mediterranean Electrotechnical Conference （MELECON） （pp. 1-5）. IEEE.
Howard, J., & Ruder, S. （2018）. Universal language model fine-tuning for text classification. arXiv preprint arXiv:1801.06146.
Kaveh, H., & Alhajj, R. （2024）. Recent advances in crack detection technologies for structures: a survey of 2022-2023 literature. Frontiers in Built Environment, 10, 1321634.
Lau, S. L., Chong, E. K., Yang, X., & Wang, X. （2020）. Automated pavement crack segmentation using u-net-based convolutional neural network. Ieee Access, 8, 114892-114899.
Lei, X., Liu, C., Li, L., & Wang, G. (2020). Automated pavement distress detection and deterioration analysis using street view map. Ieee Access, 8, 76163-76172.
Lin, T. Y., Dollár, P., Girshick, R., He, K., Hariharan, B., & Belongie, S. （2017）. Feature pyramid networks for object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition （pp. 2117-2125）.
Liu, Z., Lin, Y., Cao, Y., Hu, H., Wei, Y., Zhang, Z., ... & Guo, B. （2021）. Swin transformer: Hierarchical vision transformer using shifted windows. In Proceedings of the IEEE/CVF international conference on computer vision （pp. 10012-10022）.
Loshchilov, I., & Hutter, F. （2017）. Decoupled weight decay regularization. arXiv preprint arXiv:1711.05101.
Majidifard, H., Jin, P., Adu-Gyamfi, Y., & Buttlar, W. G. （2020）. Pavement Image Datasets: A New Benchmark Dataset to Classify and Densify Pavement Distresses. Transportation Research Record, 2674（2）, 328-339. https://doi.org/10.1177/0361198120907283 （Original work published 2020）
Majidifard, H., Adu-Gyamfi, Y., & Buttlar, W. G. (2020). Deep machine learning approach to develop a new asphalt pavement condition index. Construction and building materials, 247, 118513.
Milletari, F., Navab, N., & Ahmadi, S. A. （2016, October）. V-net: Fully convolutional neural networks for volumetric medical image segmentation. In 2016 fourth international conference on 3D vision （3DV） （pp. 565-571）. Ieee.
Mulyanto, A., Sari, R. F., Muis, A., & Harwahyu, R. (2025). Vision-Based Automated Pavement Distress Inspection: A Review. IEEE Access.
Oliveira, H., & Correia, P. L. （2009, August）. Automatic road crack segmentation using entropy and image dynamic thresholding. In 2009 17th European signal processing conference （pp. 622-626）. IEEE.
Qureshi, W. S., Hassan, S. I., McKeever, S., Power, D., Mulry, B., Feighan, K., & O’Sullivan, D. (2022). An exploration of recent intelligent image analysis techniques for visual pavement surface condition assessment. Sensors, 22(22), 9019.
Ravi, R., Bullock, D., & Habib, A. (2020). Highway and airport runway pavement inspection using mobile lidar. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 43, 349-354.
Ruseruka, C., Mwakalonge, J., Comert, G., Siuhi, S., & Perkins, J. Road condition monitoring using vehicle built-in cameras and GPS Sensors: a deep learning approach. Vehicles 5 （3）, 931–948 （2023）.
Sudre, C. H., Li, W., Vercauteren, T., Ourselin, S., & Jorge Cardoso, M. （2017, September）. Generalised dice overlap as a deep learning loss function for highly unbalanced segmentations. In International Workshop on Deep Learning in Medical Image Analysis （pp. 240-248）. Cham: Springer International Publishing.
Sun, Y., Salari, E., & Chou, E. （2009, June）. Automated pavement distress detection using advanced image processing techniques. In 2009 IEEE International Conference on Electro/Information Technology （pp. 373-377）. IEEE.
Tsai, Y., Jiang, C., & Wang, Z. (2012, January). Pavement crack detection using high-resolution 3D line laser imaging technology. In 7th RILEM International Conference on Cracking in Pavements: Mechanisms, Modeling, Testing, Detection and Prevention Case Histories (pp. 169-178). Dordrecht: Springer Netherlands.
Xiao, T., Liu, Y., Zhou, B., Jiang, Y., & Sun, J. （2018）. Unified perceptual parsing for scene understanding. In Proceedings of the European conference on computer vision （ECCV） （pp. 418-434）.
Yu, X., & Salari, E. (2011, May). Pavement pothole detection and severity measurement using laser imaging. In 2011 IEEE International Conference on Electro/Information Technology (pp. 1-5). IEEE.
Zhang, A. A., Shang, J., Li, B., Hui, B., Gong, H., Li, L., ... & Cheng, H. (2024). Intelligent pavement condition survey: overview of current researches and practices. Journal of road engineering, 4(3), 257-281.
Zhang, D., Zou, Q., Lin, H., Xu, X., He, L., Gui, R., & Li, Q. (2018). Automatic pavement defect detection using 3D laser profiling technology. Automation in Construction, 96, 350-365.
Zhang, H., Dong, Y., Hou, Y., Cheng, X., Xie, P., & Di, K. (2025). Research on Asphalt Pavement Surface Distress Detection Technology Coupling Deep Learning and Object Detection Algorithms. Infrastructures, 10(4), 72.
Zhang, Y., & Zhang, L. （2024）. Detection of pavement cracks by deep learning models of transformer and UNet. IEEE Transactions on Intelligent Transportation Systems.
Zhao, H., Qin, G., & Wang, X. （2010, October）. Improvement of canny algorithm based on pavement edge detection. In 2010 3rd international congress on image and signal processing （Vol. 2, pp. 964-967）. IEEE.
Zhao, H., Shi, J., Qi, X., Wang, X., & Jia, J. （2017）. Pyramid scene parsing network. In Proceedings of the IEEE conference on computer vision and pattern recognition （pp. 2881-2890）.