CR-YOLO:一种改进的隧道表观裂缝检测网络模型

CR-YOLO: an improved network model for crack detection in tunnel linings

摘要

针对传统裂缝检测方法复杂和泛化能力弱的缺点,提出一种基于深度学习的隧道衬砌裂缝检测网络CR-YOLO。考虑到裂缝细长的形态,该网络引入上下文引导模块(context guided block, CGBlock)融合局部与全局信息,其在局部和全局层面上结合通道和空间注意力能够使得模型充分学习到裂缝区域的上下文信息。同时添加重参数化泛化特征金字塔网络模块(reparameterized generalized feature pyramid network, RepGFPN)改善用于目标检测的特征金字塔网络(feature pyramid network, FPN),更高效地融合多尺度特征,提高对高层语义和低层空间细节的捕捉。该方法优化了计算资源下的性能,在不显著增加计算需求的前提下,减少了延迟。在自采集数据集上的试验结果表明, CR-YOLO的A_P50和A_P50-95分别可以达到90.2%和66.4%,相比于基线模型YOLOv10提高了6.2%和3.7%,在精度上优于YOLOv3、YOLOv5、YOLOv9等其他单阶段目标检测网络,推理速度达到138.7 fps,可以实现隧道衬砌裂缝实时检测。

Abstract

To address the drawbacks of cumbersome procedures and poor generalization ability in traditional crack detection methods, a CR-YOLO network for tunnel lining crack detection based on deep learning was proposed. Considering the slender morphological characteristics of cracks, the context guided block (CGBlock) was incorporated into the network to fuse local and global information; by integrating channel and spatial attention mechanisms at both local and global levels, the model was enabled to fully capture the contextual information of crack regions. Meanwhile, a reparameterized generalized feature pyramid network (RepGFPN) module was added to improve the feature pyramid network (FPN) for object detection, which achieved more efficient fusion of multi-scale features and enhanced the capture of high-level semantic information and low-level spatial details. This method optimized the model performance under constrained computing resources and reduced inference latency without a significant increase in computational overhead. Experimental results on the self-collected dataset demonstrated that the AP₅₀ and AP_50-95 of CR-YOLO reached 90.2% and 66.4%, respectively, which represented an increase of 6.2% and 3.7% compared with the baseline model YOLOv10. The model outperformed other one-stage object detection networks (YOLOv3, YOLOv5, and YOLOv9) in terms of detection accuracy. Additionally, its inference speed reached 138.7 fps, enabling real-time detection of tunnel lining cracks.

关键词

公路隧道 / 裂缝检测 / CGBlock / RepGFPN / YOLOv10 / 自动巡检

Key words

highway tunnel / crack detection / CGBlock / RepGFPN / YOLOv10 / automatic inspection and patrolling

参考文献

[1] 宋战平, 杨子凡, 张玉伟, 等. 深度学习在隧道与地下工程中的应用现状及展望[J]. 隧道建设(中英文), 2025, 45(2): 221-255.

[2] SONG Zhanping, YANG Zifan, ZHANG Yuwei, et al. Application status and prospects of deep learning in tunnels and underground engineering[J]. Tunnel Construction, 2025, 45(2): 221-255.

[3] 王晓明, 刘斌. 隧道混凝土结构病害成因分析及其健康诊断研究[J]. 混凝土, 2006(7): 89-91.

[4] WANG Xiaoming, LIU Bin. Analysis of the disease cause of formation and health diagnosing in highway tunnel concrete lining[J]. Concrete, 2006(7): 89-91.

[5] 朱合华, 潘柄屹, 武威, 等. 岩体结构面信息采集及识别方法研究进展[J]. 应用基础与工程科学学报, 2023, 31(6): 1339-1360.

[6] ZHU Hehua, PAN Bingyi, WU Wei, et al. Review on collection and extraction methods of rock mass discontinuity information[J]. Journal of Basic Science and Engineering, 2023, 31(6): 1339-1360.

[7] 张运波, 雷明锋, 肖勇卓, 等. 隧道掌子面节理卷积神经网络智能识别方法[J]. 中国公路学报, 2024, 37(7): 35-45.

[8] ZHANG Yunbo, LEI Mingfeng, XIAO Yongzhuo, et al. Intelligent recognition method of tunnel face joints and fissures using convolutional neural network[J]. China Journal of Highway and Transport, 2024, 37(7): 35-45.

[9] 李轶惠, 许振浩, 潘东东, 等. 基于数字图像的隧道岩体裂隙智能识别与参数提取方法[J]. 应用基础与工程科学学报, 2023, 31(6): 1427-1443.

[10] LI Yihui, XU Zhenhao, PAN Dongdong, et al. An intelligent identification and parameter extraction method for rock fractures based on digital images[J]. Journal of Basic Science and Engineering, 2023, 31(6): 1427-1443.

[11] 邓若曦, 王展青. 基于小波变换的路面破损图像降噪增强算法[J]. 武汉理工大学学报(信息与管理工程版), 2013, 35(5): 678-681.

[12] DENG Ruoxi, WANG Zhanqing. An image denoise algorithm of road crack based on wavelet transform[J]. Journal of Wuhan University of Technology (Information & Management Engineering), 2013, 35(5): 678-681.

[13] 王纪武, 鱼鹏飞, 罗海保. 基于改进Faster R—CNN+ZF模型的铁路桥梁裂缝分类方法[J]. 北京交通大学学报, 2020, 44(1): 106-112.

[14] WANG Jiwu, YU Pengfei, LUO Haibao. Classification method of railway bridge cracks based on improved Faster R—CNN + ZF model[J]. Journal of Beijing Jiaotong University, 2020, 44(1): 106-112.

[15] 赵凡, 李琳芸, 魏仁杰, 等. 基于通用目标检测器的大坝裂缝检测方法[J]. 数据采集与处理, 2022, 37(2): 405-414.

[16] ZHAO Fan, LI Linyun, WEI Renjie, et al. Dam crack detection method based on universal target detector[J]. Journal of Data Acquisition & Processing, 2022, 37(2): 405-414.

[17] 陈小娥. YOLO算法在交通监控目标检测中的研究与应用[J]. 智能计算机与应用, 2020, 10(6): 199-201.

[18] CHEN Xiaoe. Research and application of YOLO in traffic monitoring target detection[J]. Intelligent Computer and Applications, 2020, 10(6): 199-201.

[19] LIU Y Q, YEO H J K W. Robust pixel—wise concrete crack segmentation and properties retrieval using image patches[J]. Automation in Construction, 2021, 123: 103535.

[20] 马冬梅, 李悦媛, 陈曦. 融合双重极化注意力的轻量化半监督语义分割[J]. 计算机工程与应用, 2024, 60(8): 225-233.

[21] MA Dongmei, LI Yueyuan, CHEN Xi. Lightweight semi—supervised semantic segmentation algorithm based on dual—polarization self—attention[J]. Computer Engineering and Applications, 2024, 60(8): 225-233.

[22] 贺天宇, 祁璐帆, 李政, 等 . 基于残差网络和Faster R—CNN的混凝土裂缝智能识别[J]. 公路, 2025, 70(12): 326-331.

[23] 蔡静娴, 毕江海, 王继军, 等. 基于改进Faster R—CNN的轨旁设备识别算法研究[J]. 铁道科学与工程学报, 2022, 19(10): 3107-3116.

[24] CAI Jingxian, BI Jianghai, WANG Jijun, et al. Research on trackside equipment recognition algorithm based on improved Faster R—CNN[J]. Journal of Railway Science and Engineering, 2022, 19(10): 3107-3116.

[25] 李世威, 冯静文, 李国华, 等. 基于轻量化注意力网络的钢轨扣件状态检测方法研究[J]. 铁道科学与工程学报, 2026, 23(5): 2493-2505.

[26] LI Shiwei, FENG Jingwen, LI Guohua, et al. Rail fastener state detection using a lightweight attention network[J]. Journal of Railway Science and Engineering, 2026, 23(5): 2493-2505.

[27] 郭烊君, 雷景生. 基于改进SSD的自然场景小交通标志检测[J]. 计算机应用与软件, 2024, 41(5): 153-157.

[28] GUO Yangjun, LEI Jingsheng. Small traffic sign detection in natural scene based on improved SSD[J]. Computer Applications and Software, 2024, 41(5): 153-157.

[29] 张传伟, 李妞妞, 岳向阳, 等. 基于改进YOLOv2算法的交通标志检测[J]. 计算机系统应用, 2020, 29(6): 155-162.

[30] ZHANG Chuanwei, LI Niuniu, YUE Xiangyang, et al. Traffic sign recognition based on improved YOLOv2 algorithm[J]. Computer Systems & Applications, 2020, 29(6): 155-162.

[31] 吴江涛, 李英杰. 隧道施工环境中障碍物的轻量化目标检测算法[J]. 隧道与地下工程灾害防治, 2025, 7(1): 48-56.

[32] WU Jiangtao, LI Yingjie. The lightweight object detection algorithm for obstacles in tunnel construction environments[J]. Hazard Control in Tunnelling and Underground Engineering, 2025, 7(1): 48-56.

[33] 何铁军, 李华恩. 基于改进YOLOv5的路面病害检测模型[J]. 土木工程学报, 2024, 57(2): 96-106.

[34] HE Tiejun, LI Huaen. Pavement damage detection model based on improved YOLOv5[J]. China Civil Engineering Journal, 2024, 57(2): 96-106.

[35] 方力涵, 张庆文, 李卫国, 等. 基于改进YOLOv5的隧道衬砌病害精确检测方法[J]. 科学技术与工程, 2025, 25(16): 6812-6820.

[36] FANG Lihan, ZHANG Qingwen, LI Weiguo, et al. Tunnel lining defects precision detection method based on improved YOLOv5[J]. Science Technology and Engineering, 2025, 25(16): 6812-6820.

[37] 田佳伟, 唐子山. 基于边缘计算和ST—YOLO的矿井智能监控技术研究[J]. 煤炭工程, 2024, 56(7): 165-173.

[38] TIAN Jiawei, TANG Zishan. Mine intelligent monitoring technology based on edge computing and ST—YOLO[J]. Coal Engineering, 2024, 56(7): 165-173.

[39] 何兆益, 常宝霞, 吴逸飞, 等. 基于YOLOv5—IBX网络模型的公路隧道衬砌裂缝检测方法研究[J]. 沈阳建筑大学学报(自然科学版), 2023, 39(5): 888-898.

[40] HE Zhaoyi, CHANG Baoxia, WU Yifei, et al. Research on crack detection method of highway tunnel lining based on YOLOv5—IBX network model[J]. Journal of Shenyang Jianzhu University (Natural Science), 2023, 39(5): 888-898.

[41] 贺泳超, 陈秋南, 程家杰 . 基于改进YOLOv7算法的隧道衬砌裂缝智能识别[J]. 湖南科技大学学报(自然科学版), 2024, 39(2): 35-43.

[42] 寇俊伟, 李艳萍, 姚剑, 等. 基于改进YOLOv5的隧道衬砌裂缝检测方法[J]. 中国铁路, 2025(5): 96-104.

[43] KOU Junwei, LI Yanping, YAO Jian, et al. Tunnel lining crack inspection method based on improved YOLOv5[J]. China Railway, 2025(5): 96-104.

[44] 胥铁峰, 黄河, 张红民, 等. 基于改进YOLOv8的轻量化道路病害检测方法[J]. 计算机工程与应用, 2024, 60(14): 175-186.

[45] XU Tiefeng, HUANG He, ZHANG Hongmin, et al. Lightweight road damage detection method based on improved YOLOv8[J]. Computer Engineering and Applications, 2024, 60(14): 175-186.

[46] 张振海, 孙岩, 李哲远. 基于MDS—YOLO的轻量级隧道表观病害检测算法[J]. 交通运输工程学报, 2025, 25(6): 271-283.

[47] ZHANG Zhenhai, SUN Yan, LI Zheyuan. Lightweight tunnel surface defect detection algorithm based on MDS—YOLO[J]. Journal of Traffic and Transportation Engineering, 2025, 25(6): 271-283.

[48] 朱佳慧, 刘艺, 张登银. 基于改进YOLOv8n的道路裂缝检测轻量化模型[J]. 数据采集与处理, 2025, 40(5): 1333-1347.

[49] ZHU Jiahui, LIU Yi, ZHANG Dengyin. A lightweight road crack detection model based on improved YOLOv8n[J]. Journal of Data Acquisition and Processing, 2025, 40(5): 1333-1347.

[50] 张亚军, 苗皓源, 马薇, 等. 基于YOLOv8改进的无人机航拍路面损伤检测算法[J]. 电子测量技术, 2026, 49(2): 181-191.

[51] ZHANG Yajun, MIAO Haoyuan, MA Wei, et al. YOLOv8—based improved algorithm for road damage detection in UAV aerial images[J]. Electronic Measurement Technology, 2026, 49(2): 181-191.

[52] 董延东, 李强, 王剑豪. 基于改进YOLOv8s的隧道衬砌裂缝的实例分割算法[J]. 计算机应用, 2025, 45(增刊1): 263-270.

[53] DONG Yandong, LI Qiang, WANG Jianhao. Instance segmentation algorithm for tunnel lining cracks based on improved YOLOv8s[J]. Journal of Computer Applications, 2025, 45(Suppl.1): 263-270.

[54] 彭雨诺, 刘敏, 万智, 等. 基于改进YOLO的双网络桥梁表观病害快速检测算法[J]. 自动化学报, 2022, 48(4): 1018-1032.

[55] PENG Yunuo, LIU Min, WAN Zhi, et al. A dual deep network based on the improved YOLO for fast bridge surface defect detection[J]. Acta Automatica Sinica, 2022, 48(4): 1018-1032.

[56] 朱前坤, 谢辰辉, 张琼, 等. 基于计算机视觉和深度学习的古桥裂缝识别方法[J]. 西南交通大学学报, 2026, 61(2): 529-540.

[57] ZHU Qiankun, XIE Chenhui, ZHANG Qiong, et al. A method for crack detection in ancient bridges based on computer vision and deep learning[J]. Journal of Southwest Jiaotong University, 2026, 61(2): 529-540.

[58] 杨豪, 郭凯丽, 张军辉, 等. 基于YOLO11—L算法的道路内部空洞定量识别[J/OL]. 中国公路学报, 2025(2025—08—18)[2026—01—18]. https://link.cnki.net/urlid/61.1313.U.20250818.1134.002.

[59] YANG Hao, GUO Kaili, ZHANG Junhui, et al. Quantitative identification of internal voids in roads based on YOLO11—L algorithm[J/OL]. China Journal of Highway and Transport, 2025(2025—08—18)[2026—01—18]. https://link.cnki.net/urlid/61.1313.U.20250818.1134.002.

[60] 鲍艳, 梅崇斌, 徐鹏宇, 等. 基于改进YOLOv8算法的地铁隧道裂缝识别方法研究[J]. 隧道建设(中英文), 2024, 44(10): 1961-1970.

[61] BAO Yan, MEI Chongbin, XU Pengyu, et al. Crack identification in metro tunnels based on improved YOLOv8 algorithm[J]. Tunnel Construction, 2024, 44(10): 1961-1970.

[62] 董明书, 陈俐企, 马川义, 等. 沥青路面内部裂缝雷达图像智能判识算法研究[J]. 山东大学学报(工学版), 2025, 55(3): 72-79.

[63] DONG Mingshu, CHEN Liqi, MA Chuanyi, et al. Deep learning—based intelligent judgment for radar detection of pavement cracks[J]. Journal of Shandong University (Engineering Science), 2025, 55(3): 72-79.

[64] 英红, 覃云涛, 刘曦民, 等. 基于改进YOLOv8的沥青路面深度图像病害检测算法[J]. 湖南科技大学学报(自然科学版), 2025, 40(1): 88-101.

[65] YING Hong, QIN Yuntao, LIU Ximin, et al. Asphalt pavement depth image disease detection algorithm based on improved YOLOv8[J]. Journal of Hunan University of Science and Technology (Natural Science Edition), 2025, 40(1): 88-101.

[66] 陈智丽, 张伍彪, 王冰, 等. Ghost—YOLO:复杂环境下混凝土结构裂缝病害检测网络[J]. 计算机应用与软件, 2025, 42(2): 171-180.

[67] CHEN Zhili, ZHANG Wubiao, WANG Bing, et al. Ghost—YOLO: crack disease detection network of concrete structure in complex environments[J]. Computer Applications and Software, 2025, 42(2): 171-180.

[68] 武斌, 于双玲, 陈杨杨, 等. 结合三维无参数注意力机制的隧道裂缝检测方法[J]. 隧道建设(中英文), 2024, 44(7): 1520-1531.

[69] WU Bin, YU Shuangling, CHEN Yangyang, et al. Tunnel crack detection method based on three—dimensional parameter—free attention mechanisms[J]. Tunnel Construction, 2024, 44(7): 1520-1531.

[70] 杨琚钱, 胡平, 戴家树. 基于Yolov8—scG神经网络的电动车头盔佩戴检测算法[J]. 重庆工商大学学报(自然科学版), 2026, 43(2): 101-107.

[71] YANG Juqian, HU Ping, DAI Jiashu. An electric vehicle helmet wear detection algorithm based on Yolov8—scG neural network[J]. Journal of Chongqing Technology and Business University (Natural Science Edition), 2026, 43(2): 101-107.

[72] WANG A, CHEN H, LIU L H, et al. YOLOv10: real—time end—to—end object detection[C]// Advances in Neural Information Processing Systems 37 (NeurIPS 2024). Vancouver, Canada: Neural Information Processing Systems Foundation, Inc. (NeurIPS), 2024: 107984-108011.

[73] 陈森森, 程刚, 王龙腾. 基于YOLO—RepGFPN模型的煤矸检测方法研究[J]. 佳木斯大学学报(自然科学版), 2024, 42(3): 6-11.

[74] CHEN Sensen, CHENG Gang, WANG Longteng. Research on coal gangue detection method based on YOLO—RepGFPN model[J]. Journal of Jiamusi University (Natural Science Edition), 2024, 42(3): 6-11.

[75] 肖剑, 杨小苑, 何昕泽, 等. 基于全局信息感知的轻量级螺纹钢表面缺陷检测算法[J]. 浙江大学学报(工学版), 2026, 60(7): 1438-1451.

[76] XIAO Jian, YANG Xiaoyuan, HE Xinze, et al. Lightweight rebar surface defect detection algorithm based on global information perception[J]. Journal of Zhejiang University (Engineering Science), 2026, 60(7): 1438-1451.

[77] 方恩权, 王耀东, 李星言, 等. 地铁盾构隧道渗漏水病害图像识别算法[J]. 隧道与地下工程灾害防治, 2022, 4(4): 28-33.

[78] FANG Enquan, WANG Yaodong, LI Xingyan, et al. Algorithms of leakage disease images recognition in subway tunnel[J]. Hazard Control in Tunnelling and Underground Engineering, 2022, 4(4): 28-33.

[79] ROY A M, BHADURI J. DenseSPH—YOLOv5: an automated damage detection model based on DenseNet and Swin—Transformer prediction head—enabled YOLOv5 with attention mechanism[J]. Advanced Engineering Informatics, 2023, 56: 102007.

[80] WANG C Y, YEH I H, MARK LIAO H Y. YOLOv9: learning what you want to learn using programmable gradient information[C]// Computer Vision—ECCV 2024. Milan, Italy: Springer, 2025: 1-21.

基金资助

山东高速创新联合基金资助项目(Y110061Q01)

中国铁道科学研究院集团有限公司科研开发计划重大资助项目(2023YJ027)