定期檢測對橋梁而言是基礎且重要的措施，可以幫助我們及時瞭解橋梁劣化的產生以及防止劣化擴散。傳統的橋檢方式是由橋檢技師到現場，以人工目視的方式去執行。必要時，可能需要使用繩索或是高空作業平台等儀器輔助進行，是一項耗時且勞力密集型的任務。相比之下，若是能使用無人機或是相機等儀器取得近距離的橋梁影像，就有機會達到自動化的橋梁劣化檢測。為了評估自動化辨識影像中劣化的可行性，本研究修改了神經網路模型DeepLab V3+，比較了不同骨幹網路於特徵萃取的效果以及調整損失函數對於辨識結果的差異。
最終，我們將DeepLab V3+的骨幹網路替換為DRN模型，並且使用一結合了focal loss function與hard example term的損失函數-FL_HE進行辨識。我們提出了這個自動化的橋梁劣化檢測方法-DeepLab V3++，並且提升了模型的辨識能力及效率。在此研究中，我們使用了橋檢技師於過往的檢測任務中蒐集的，上傳於台灣橋梁管理系統(Taiwan Bridge Management System 2, TBMS-2) 的影像進行訓練。不同於其他研究，本研究使用的是包含多種劣化且背景複雜的影像。在橋檢技師及土木系專家的指導下，我們參考台灣「公路橋梁檢測及補強規範」，人工標註6種橋梁劣化類型，分別為裂縫、鏽蝕、剝落、白華、滲水以及銹水。在驗證資料集的部分，另外使用了14座橋梁影像進行辨識成果的比較及分析。結果顯示，本研究提出的DeepLab V3++ 在6種橋梁劣化類型的辨識皆比原始的DeepLab V3+ 準確。在精度評估的部分，各項劣化類別的交並比(Intersection over Union, IoU)皆有顯著的提升，且有效降低漏授誤差(Omission Error, OE)及誤授誤差(Commission Error, CE)。此方法提升了模型對於橋梁劣化的整體辨識能力。在本研究中，我們展示了深度學習模型於輔佐橋檢技師進行橋梁劣化辨識的潛力，能夠協助他們以更高效率、低風險的方式完成橋檢任務。

Regular deterioration detection is a fundamental and crucial task for bridges; it can bring people a timely understanding of the generation of deterioration and prevent defect expansion. The traditional bridge deterioration detection is implemented by on-site inspection of bridge technicians. They may require the ropes and Aerial Work Platforms (AWPs) for detection tasks if necessary. It’s a time-consuming and labor-intensive task. By contrast, it would be possible to execute automatic bridge maintenance tasks if they could use Unmanned Aerial Vehicles (UAVs) or phones to acquire close-range images for deterioration evaluation. This research modified the neural network DeepLab V3+, compared the effects of using different backbone networks for feature extraction, and the discrepancy of adjusting the loss function and recognition results.
At last, we use DRN to replace the original backbone network of DeepLab V3+ and combine it with a modified loss function – FL_HE, which is the focal loss function and added with a hard example term. We proposed an automatic bridge deterioration detection method - DeepLab V3++ and enhanced the recognition ability and efficiency. The concrete bridge images on TBMS-2 (Taiwan Bridge Management System 2) collected from the past bridge inspection are used for model training. Unlike other research, we use real-world images that contain multiple deteriorations and complicated background conditions. Follow the instructions of bridge technicians and experts in the civil engineering department and refer to “Highway Bridge Inspection and Reinforcement Specification”; we manually label 6 categories of deterioration, including cracks, corrosion, spalling, efflorescence, infiltration, and rusty stain. We also use the images from 14 unseen bridges as validation data and quantify the recognition results for comparison. The results prove that the proposed method DeepLab V3++ outperforms the original DeepLab V3+; It works better for 6 defect detection. DeepLab V3++ shows significant improvement in Intersection over Union (IoU) and reduces Commission Error (CE) and Omission Error (OE). The proposed method enhances the overall recognition results. In this research, the deep learning method has the makings of reaching automatic bridge deterioration detection, assisting the bridge technicians in finishing the inspection tasks, lowering the risks, and enhancing the implementation efficiency.

Badrinarayanan, V., Kendall, A., & Cipolla, R. (2017). Segnet: A deep convolutional encoder-decoder architecture for image segmentation. IEEE transactions on pattern analysis and machine intelligence, 39(12), 2481-2495.
Brooks, C., Cook, C., Dobson, R., Oommen, T., Zhang, K., Mukherjee, A., Samsami, R., Semenchuk, A., Lovelace, B., & Hung, V. (2022). Integration of Unmanned Aerial Systems Data Collection Into Day-to-Day Usage for Transportation Infrastructure–A Phase III Project.
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.
Chen, L.-C., Papandreou, G., Schroff, F., & Adam, H. (2017). Rethinking atrous convolution for semantic image segmentation. arXiv preprint arXiv:1706.05587.
Chen, L.-C., Zhu, Y., Papandreou, G., Schroff, F., & Adam, H. (2018). Encoder-decoder with atrous separable convolution for semantic image segmentation. Proceedings of the European conference on computer vision (ECCV),
Chollet, F. (2017). Xception: Deep learning with depthwise separable convolutions. Proceedings of the IEEE conference on computer vision and pattern recognition,
Dong, Y., Wang, J., Wang, Z., Zhang, X., Gao, Y., Sui, Q., & Jiang, P. (2019). A deep-learning-based multiple defect detection method for tunnel lining damages. IEEE Access, 7, 182643-182657.
He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. Proceedings of the IEEE conference on computer vision and pattern recognition,
Hinton, G. E., Osindero, S., & Teh, Y.-W. (2006). A fast learning algorithm for deep belief nets. Neural computation, 18(7), 1527-1554.
Howard, A. G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T., Andreetto, M., & Adam, H. (2017). Mobilenets: Efficient convolutional neural networks for mobile vision applications. arXiv preprint arXiv:1704.04861.
König, J., Jenkins, M. D., Mannion, M., Barrie, P., & Morison, G. (2021). Optimized deep encoder-decoder methods for crack segmentation. Digital Signal Processing, 108, 102907.
Kim, I.-H., Jeon, H., Baek, S.-C., Hong, W.-H., & Jung, H.-J. (2018). Application of crack identification techniques for an aging concrete bridge inspection using an unmanned aerial vehicle. Sensors, 18(6), 1881.
Kirillov, A., He, K., Girshick, R., Rother, C., & Dollár, P. (2019). Panoptic segmentation. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition,
Kruachottikul, P., Cooharojananone, N., Phanomchoeng, G., Chavarnakul, T., Kovitanggoon, K., Trakulwaranont, D., & Atchariyachanvanich, K. (2019). Bridge sub structure defect inspection assistance by using deep learning. 2019 IEEE 10th International Conference on Awareness Science and Technology (iCAST),
LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. nature, 521(7553), 436-444. https://doi.org/10.1038/nature14539
Li, K., Wang, B., Tian, Y., & Qi, Z. (2021). Fast and accurate road crack detection based on adaptive cost-sensitive loss function. IEEE Transactions on Cybernetics.
Lin, T.-Y., Goyal, P., Girshick, R., He, K., & Dollár, P. (2017). Focal loss for dense object detection. Proceedings of the IEEE international conference on computer vision,
Liu, W., Wang, Z., Liu, X., Zeng, N., Liu, Y., & Alsaadi, F. E. (2017). A survey of deep neural network architectures and their applications. Neurocomputing, 234, 11-26. https://doi.org/10.1016/j.neucom.2016.12.038
Narazaki, Y., Hoskere, V., Hoang, T. A., & Spencer Jr, B. F. (2018). Automated vision-based bridge component extraction using multiscale convolutional neural networks. arXiv preprint arXiv:1805.06042.
Qu, Z., Cao, C., Liu, L., & Zhou, D.-Y. (2021). A deeply supervised convolutional neural network for pavement crack detection with multiscale feature fusion. IEEE transactions on neural networks and learning systems.
Radovic, M., Adarkwa, O., & Wang, Q. (2017). Object recognition in aerial images using convolutional neural networks. Journal of Imaging, 3(2), 21.
Ronneberger, O., Fischer, P., & Brox, T. (2015). U-net: Convolutional networks for biomedical image segmentation. Medical Image Computing and Computer-Assisted Intervention–MICCAI 2015: 18th International Conference, Munich, Germany, October 5-9, 2015, Proceedings, Part III 18,
Shrivastava, A., Gupta, A., & Girshick, R. (2016). Training region-based object detectors with online hard example mining. Proceedings of the IEEE conference on computer vision and pattern recognition,
Song, W., Jia, G., Zhu, H., Jia, D., & Gao, L. (2020). Automated pavement crack damage detection using deep multiscale convolutional features. Journal of Advanced Transportation, 2020.
Sun, Y., Yang, Y., Yao, G., Wei, F., & Wong, M. (2021). Autonomous crack and bughole detection for concrete surface image based on deep learning. IEEE Access, 9, 85709-85720.
Xu, H., Su, X., Xu, H., & Li, H. (2019). Autonomous bridge crack detection using deep convolutional neural networks. 3rd International Conference on Computer Engineering, Information Science & Application Technology (ICCIA 2019),
Yu, F., Koltun, V., & Funkhouser, T. (2017). Dilated residual networks. Proceedings of the IEEE conference on computer vision and pattern recognition,
Yun, S., Han, D., Oh, S. J., Chun, S., Choe, J., & Yoo, Y. (2019). Cutmix: Regularization strategy to train strong classifiers with localizable features. Proceedings of the IEEE/CVF international conference on computer vision,
Zhao, H., Shi, J., Qi, X., Wang, X., & Jia, J. (2017). Pyramid scene parsing network. Proceedings of the IEEE conference on computer vision and pattern recognition,
Zou, Q., Zhang, Z., Li, Q., Qi, X., Wang, Q., & Wang, S. (2018). Deepcrack: Learning hierarchical convolutional features for crack detection. IEEE Transactions on Image Processing, 28(3), 1498-1512.
王炤烈, 宋裕祺, 林曜滄, 彭康瑜, & 黃炳勳. (2018). Morandi 橋崩塌帶來的省思與建議. 土木水利, 45(5), 44-45.
姚乃嘉, 蔡閔光, 黃榮堯, 葉啟章, 許文科, 任以永, 廖先格, 莊友涵, & 廖艾貞. (2015). 第二代臺灣地區橋梁管理資訊系統建置規劃 [交通部運輸研究所計畫報告].
彭康瑜, 林曜滄, 宋裕祺, 洪曉慧, 蘇進國, & 吳明遠. (2018). Morandi 橋崩塌事件分析報導. 土木水利, 45(5), 5-21.
曾子峻. (2011). 多軸飛行器於橋梁檢測之應用研究。 國立高雄應用科技大學].
薛強, 張權, 翁健煌, 徐偉誌, 陳正忠, & 葉啟章. (2012). 台灣橋梁檢測有效實施建議. 中興工程(117), 31-40.