排水管道的健康状况影响着城市排水系统的正常运行。现多采用人工判读CCTV检测图像的方法多存在效率低下、误检率高等问题，因此开展排水管道缺陷的智能检测与识别研究具有重要现实意义。为辅助工程人员快速准确的完成排水管道检测，论文基于深度学习理论，系统研究了排水管道缺陷检测及识别算法，取得了以下研究成果。

针对人工缺陷检测方法准确率低、类型识别不全的问题，提出一种优化YOLO v4模型的排水管道缺陷检测与识别算法(YOLO-sewer)。为了提升模型对关键特征信息的捕获能力，算法融合SE通道注意力机制改进了YOLO v4残差单元，构建基于残差注意力单元的残差注意力主干网络。同时引入RFB感受野增强模块扩增感受野范围并减少计算量。实验结果表明：相较于SSD、YOLO v3、YOLO v4等检测算法，YOLO-sewer的平均检测准确率分别提升了22.01%和8.83%和7.54%，对于部分小样本缺陷的检测效果也实现了大幅度提升，实现了对沉积、错口等14种常见排水管道缺陷的检测与有效识别。

针对管道缺陷检测中，小样本、细粒度缺陷目标识别效果不佳的问题，提出一种改进Transformer的排水管道缺陷精确识别算法(LPSST)。算法以Swin Transformer模型为改进的主干网，在浅层网络中新增局部区域建议模块，进一步提高模型对局部特征信息的捕获能力。LPSST使用融合伪标签和一致性正则化的半监督训练方法提升算法泛化性，对改进模型强化训练发挥视觉Transformer架构大体量数据集训练的优势。实验结果表明：相较于Resnet、Efficentnet等其他识别算法，LPSST取得了最佳识别效果，识别准确率达92.65%，超过原始主干网17.25%。算法实现了对错口、变形等四种特征相似性管道结构性缺陷的精确识别。

The health status of drainage pipes affects the normal operation of urban drainage systems. Most of the current mainstream detection methods have some problems, such as low efficiency and a high error detection rate, so it is of great practical significance to carry out intelligent detection and identification of drainage pipeline defects. To assist engineers in completing drainage pipeline detection quickly and accurately, this paper systematically studies the types of drainage pipeline defects and related detection and recognition algorithms based on deep learning theory and has obtained important research results.

Aiming at the problems of low detection accuracy and incomplete recognition types of manual defect detection, this paper proposes a drainage pipeline defect detection and recognition algorithm (YOLO-sewer), which optimizes the YOLO v4 model. The algorithm is integrates SE channel attention to improved YOLO v4 residual unit, and build a residual attention backbone network based on residual attention unit to improve the model's ability to capture key feature information. Introduces the RFB receptive field enhancement module to expand the receptive field range and reduce the amount of computation. The experimental results compared with detection algorithms like SSD, YOLO v3, YOLO v4, the average detection accuracy of YOLO-sewer has increased by 22.01 % and 8.83 % and 7.54 %, respectively, and the detection effect of some small sample defects has also greatly improved. The algorithm realizes the detection and identification of 14 common drainage pipeline defects, such as deposition and stagger.

Aiming at the poor recognition effect of the drainage pipes defect detection for small samples and fine-grained defect targets, this paper proposes an improved Transformer algorithm (LPSST) for the accurate recognition of drainage pipeline defects. The algorithm takes the Swin Transformer model as the improved backbone network and adds a local area suggestion module in the shallow network to further improve the ability of the model to capture local feature information. In addition, LPSST uses a semi-supervised training method that fuses pseudo labels and consistency regularization to improve the generalization of the algorithm and strengthens the training of the improved model to give full play to the advantages of the visual Transformer architecture in training large datasets. Compared with other recognition algorithms, such as Resnet and Efficentnet, LPSST has achieved the best recognition results, with a recognition accuracy of 92.65 %, exceeding the original backbone network by 17.25 %. The experimental results show that the algorithm can accurately identify the structural defects of pipelines with four similar characteristics, such as staggered and deformation.

[1] 中华人民共和国统计局. 中国统计年鉴[M]. 北京: 中国统计出版社, 2021.

[2] B. T. Rushton. Low-impact parking lot design reduces runoff and pollutant loads[J]. Journal of Water Resources Planning & Management, 2001, 127(3): 172-179.

[3] C. Allendeprieto, B. I. Mendezfernandez, L. A. Sanudofontaneda, et al. Development of ageospatial data-based methodology for stormwater management in urban areas using ffreely-available software[J]. International Journal of Environmental Research and Public Health, 2018, 15(8): 1703-1713.

[4] 王帅超. 城市地下管道渗漏引起的路面塌陷机理分析与研究[D]. 郑州: 郑州大学, 2017.

[5] 张辰. 加强城镇排水管网规划建设管理保障高效安全运行[M]. 北京: 给水排水, 2016, 52(8): 1-3.

[6] R. Ashley, B. Crabtree, A. Fraser. Recent european research into the behavior of sewer sediments and associated pollutants and processes[C]//Building Partnerships. Reston: ASCE Press, 2014: 1-11.

[7] 侯精明, 康永德, 李轩, 等. 西安市暴雨致涝成因分析及对策[J]. 西安理工大学学报, 2020, 36(03): 269-274.

[8] CJJ 181-2012. 城镇排水管道检测与评估技术规程[S]. 北京: 中华人民共和国住房和城乡建设部, 2012.

[9] 陈希刚. 市政管道CCTV检测技术应用[J]. 水利水电施工, 2019, 38(1): 102-105.

[10] O. Huynh, R. Ross, A. Martchenko, et al. Anomaly inspection in sewer pipes using stereo vision[C]//IEEE International Conference on Signal and Image Processing Applications (ICSIPA). New York: IEEE, 2015: 60-64.

[11] 雷芳芳. CCTV技术在福州市排水管道检测中的应用研究[J]. 给水排水, 2019, 55(S1): 275-276.

[12] 肖倩, 王俊然, 陈辉, 等. 深圳市某片区排水管道CCTV检测评估与修复方案[J]. 给水排水, 2019, 55(9): 109-114．

[13] A. Hawari, M. Alamin, F. Alkadour, et al. Automated defect detection tool for closed circuit television (CCTV) inspected sewer pipelines[J]. Automation in Construction, 2018, 89: 99-109.

[14] R. Wirahadikusumah, D. Abraham, T. Iseley. Challengingissues in modeling deterioration of combined sewers[J]. Journal of Infrastructure Systems, 2001, 7(2): 77-84.

[15] D. J. Titman. Applications of thermography in non-destructive testing of structures[J]. NDT & E International, 2001, 34(2): 149-154.

[16] M. H. Siqueira, C. E. Gatts, R. R. da Silva, et al. The use of ultrasonic guided wavesand wavelets analysis in pipe inspection[J]. Ultrasonics, 2004, 41(10): 785-797.

[17] J. Zhang, B. W. Drinkwater, P. D. Wilcox. Effect of roughness on imaging and sizing rough crack-like defects using ultrasonic arrays[J]. IEEE Transactions on Ultrasonics, Ferroelectrics and Frequency Control, 2012, 59(5): 939-948.

[18] 夏娟, 黄民, 李天剑, 等. 排水管道超声波数据采集系统[J]. 北京信息科技大学学报(自然科学版), 2011, 26(3): 69-73.

[19] 陈高泉. 排水管道超声波成像检测系统研究与设计[D]. 沈阳:沈阳工业大学, 2005.

[20] 魏效国, 黄民. 超声波探测排水管道的可行性研究[J]. 机械工程师, 2016, 51(1): 4-6.

[21] 崔英良, 李海明, 王延军, 等. CCTV和探地雷达技术在城市道路病害中的应用[J]. 城市勘测, 2018, 34(S1): 141-143.

[22] 户莹. 基于深度学习的地下排水管道缺陷智能检测技术研究[D]. 西安: 西安理工大学, 2019.

[23] P. W. Fieguth, S. K. Sinha. Automated analysis and detection of cracks in underground scanned pipes[C]//Proceedings of the 1999 International Conference on Image Processing. New York: IEEE, 1999, 4: 395-399.

[24] S. K. Sinha, F. Karray. Classification of underground pipe scanned images using feature extraction and neuro-fuzzy algorithm[J]. IEEE Transactions on Neural Networks, 2002, 13(2): 393-401.

[25] 马永军, 方凯, 王定成, 等. 基于支持向量机和距离度量的管道内表面图像分类方法研究[J]. 数据采集与处理, 2002(02): 151-155.

[26] M. Yang. Segmenting ideal morphologies of sewer pipe defects on CCTV images for automated diagnosis [J]. Expert System with Applications, 2008, 02(6): 5-7.

[27] 孙文雅, 黄民, 李天剑, 等. 基于BP神经网络管道裂缝图像分割[J]. 计算机测量与控制, 2012,20(05): 1363-1364+1368.

[28] M. Motamedi, F. Faramarzi, O. Duran. New concept for corrosion inspection of urban pipeline networks by digital image processing[C]//Proceedings of the 38th Annual Conference on IEEE Industrial Electronics Society. New York: IEEE, 2012: 1551-1556.

[29] M. A. Alam, M. M. N. Ali, M. A. A. A. Syed, et al. An algorithm to detect and identify defects of industrial pipes using image processing[C]//Proceedings of the 8th International Conference on Software, Knowledge, Information Management and Applications. New York: IEEE, 2014: 1-6.

[30] 刘喆. 基于图像分析的管道缺陷特征提取方法研究[D]. 沈阳: 东北大学, 2014.

[31] D. S. Mayuri, W. Kishor. An Application of Image Processing to Detect the Defects of Industrial Pipes[J]. International Journal of Advanced Research in Computer and Communication Engineering, 2016, 5(3): 979-981.

[32] 李忠虎, 张琳, 闫俊红. 管道腐蚀视觉测量图像边缘检测算法研究[J]. 电子测量与仪器学报, 2017, 31(11): 1788-1795.

[33] 黄玉龙. 基于视频图像的管道裂纹缺陷检测方法研究[D]. 西安: 西安理工大学, 2018.

[34] 蔡文浩. 城市地下管道视觉检测关键技术及应用研究[D]. 武汉: 湖北工业大学, 2018.

[35] Z. Chen, T. C. Hutchinson. Image-based framework for concrete surface crack monitoring and quantification[J]. Advances in Civil Engineering, 2010(2): 9-26．

[36] H. W. Cho, H. J. Yoon, J. C. Yoon. Analysis of crack image recognition characteristics in concrete structures depending on the illumination and image acquisition distance through outdoor experiments[J]. Sensors, 2016, 16(10): 1646-1664.

[37] S. Sinha, P. W. Fieguth. Morphological segmentation and classification of underground pipe images[J]. Machine Vision and Applications, 2006, 17(1): 21-31.

[38] S. K. Sinha, P. W. Fieguth. Segmentation of buried concrete pipe images[J]. Automation in Construction, 2006, 15(1): 47-57.

[39] S. Iyer, S. K. Sinha. Segmentation of pipe images for crack detection in buried sewers[J]. Computer-Aided Civil and Infrastructure Engineering, 2006, 21(6): 395-410.

[40] B. Neury, M. C. Tania. Automated detection of welding defects in pipelines from radiographic images DWDI[J]. NDT & E International, 2017, 86: 7-13.

[41] S. S. Kumar, D. M. Abraham, M. R. Jahanshahi, et al. Automated defect classification in sewer closed circuit television inspections using deep convolutional neural networks[J]. Automation in Construction, 2018, 91: 273-283.

[42] S. Ren, K. He, G. Ross. et al. Faster r-cnn: Towards real-time object detection with region proposal networks[J]. Advances in neural information processing systems, 2015, 28: 91-99.

[43] J. Cheng, M. Wang. Automated detection of sewer pipe defects in closed-circuit television images using deep learning techniques[J]. Automation in Construction, 2018, 95(11): 155-171.

[44] 王庆, 姚俊, 谭文禄, 等. 基于Faster R-CNN的排水管道缺陷检测研究[J]. 软件导刊, 2019, 18(10):6.

[45] X. Yin, Y. Chen, A. Bouferguene, et al. A deep learning-based framework for an automated defect detection system for sewer pipes[J]. Automation in Construction, 2019, 109(2020): 102967.

[46] I. H. Syed, L. Minh Dang, M. Irfan, et al. Underground sewer pipe condition assessment based on convolutional neural networks[J]. Automation in Construction, 2019, 106: 102849.

[47] D. Li, A. Cong, S. Guo. Sewer damage detection from imbalanced CCTV inspection data using deep convolutional neural networks with hierarchical classification[J]. Automation in Construction, 2019, 101(5): 199-208.

[48] 周倩倩, 司徒祖祥, 腾帅, 等. 基于卷积神经网络的排水管道缺陷智能检测与分类[J]. 中国给水排水, 2021, 37(21): 114-118.

[49] 王大成, 陈国强, 秦军,等. 排水管道闭路电视视频智能检测技术的现状与挑战[J]. 工程勘察, 2022, 50(03): 52-56.

[50] 李伟, 刘桂雄, 曾成刚. 基于实例分割+CCTV排水管道缺陷检测方法研究[J]. 电子测量技术, 2022, 45(03): 153-157.

[51] 王和平, 安关峰, 谢广永. 《城镇排水管道检测与评估技术规程》(CJJ181-2012)解读[J]. 给水排水, 2014, 40(2): 124-127.

[52] R. Joseph, D. Santosh, G. Ross. et al. You only look once: Unified, real-time object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. New York: IEEE, 2016: 779-788.

[53] J. Redmon, A. Farhadi. YOLO9000: better, faster, stronger[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. New York: IEEE, 2017, 7263-7271.

[54] J. Redmon, A. Farhadi YOLOv3: An Incremental Improvement[J]. arXive-prints, 2018.

[55] B. Alexey, C. Wang, H. Mark Liao. Yolo v4: Optimal speed and accuracy of object detection[J]. arXiv preprint arXiv: 2004. 10934, 2020.

[56] K. He, X. Zhang, S. Ren, et al. Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence, 2014, 37(9): 1904-16.

[57] S. Liu, L. Qi, H. Qin, et al. Path aggregation network for instance segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2018: 8759-8768.

[58] L. Wei, A Dragomir, E. Dumitru, et al. Ssd: Single shot multibox detector[C]//Proceedings of the European conference on computer vision. Germany: Springe, 2016: 21-37.

[59] S. Liu, D. Huang. Receptive field block net for accurate and fast object detection[C]//Proceedings of the European conference on computer vision. 2018: 385-400.

[60] J. Hu, L. Shen, S. Albanie, et al. Squeeze-and-Excitation Networks[J]. IEEE transactions on pattern analysis and machine intelligence, 2019, 7132-7141.

[61] S. D’Ascoli, H. Touvron, M. L. Leavitt, et al. Convit: Improving vision transformers with soft convolutional inductive biases[C]//Proceedings of the International Conference on Machine Learning. Virtual: PMLR Press, 2021: 2286-2296.

[62] A. Vaswani, N. Shazeer, N. Parmar, et al. Attention is all you need[J]. Advances in neural information processing systems, 2017, 30: 1-10.

[63] Z. Liu, Y. Lin, Y. Cao, et al. Swin transformer: Hierarchical vision transformer using shifted windows[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. New York: IEEE, 2021, 10012-10022.

[64] K. Sohn, D. Berthelot, N. Carlini, et al. Fixmatch: Simplifying semi-supervised learning with consistency and confidence[J]. Advances in Neural Information Processing Systems, 2020, 33: 596-608.

[65] T. Lin, P. Dollar, R. Girshick, et al. Feature Pyramid Networks for Object Detection[J]. IEEE Computer Society, 2017, 2117-2125.

[66] C. Wang, H. Liao, Y. Wu, et al. CSPNet: A new backbone that can enhance learning capability of CNN[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops. New York: IEEE, 2020: 390-391.