基于计算机视觉与混合测量技术的结构裂缝识别方法

李泽伟; 杨永清; 廖曼; 谢明志; 刘雨; 黄胜前

doi:10.3969/j.issn.0258-2724.20230700

基于计算机视觉与混合测量技术的结构裂缝识别方法

doi: 10.3969/j.issn.0258-2724.20230700

李泽伟^{1, 2,},
杨永清^{1, 2},
廖曼¹,
谢明志^{1, 2, ,},
刘雨¹,
黄胜前¹

1.
西南交通大学土木工程学院，四川成都 610031
2.
西南交通大学桥梁智能与绿色建造全国重点实验室，四川成都 611756

基金项目: 国家自然科学基金项目（52322811）；四川省科技计划（2020YJ0081）

详细信息

作者简介:
李泽伟（1998—），男，博士研究生，研究方向为桥梁智能检测及损伤评估，E-mail：lizw0616@163.com

通讯作者:
谢明志（1985—），男，副教授，博士，研究方向为桥梁智能检测及损伤评估，E-mail：mzxie@home.swjtu.edu.cn

中图分类号: U445
计量
- 文章访问数: 122
- HTML全文浏览量: 99
- PDF下载量: 37
- 被引次数: 0
出版历程
- 收稿日期: 2023-12-23
- 修回日期: 2024-06-21
- 网络出版日期: 2025-09-22

Structural Crack Detection Based on Computer Vision and Hybrid Measurement Technology

LI Zewei^{1, 2
,},
YANG Yongqing^{1, 2},
LIAO Man¹,
XIE Mingzhi^{1, 2
, ,},
LIU Yu¹,
HUANG Shengqian¹

1.
School of Civil Engineering, Southwest Jiaotong University, Chengdu 610031, China
2.
State Key Laboratory of Bridge Intelligent and Green Construction, Southwest Jiaotong University, Chengdu 611756, China

摘要

摘要:
混凝土表面裂缝检测为桥梁结构的运维提供关键技术资料与决策要素，然而，裂缝识别作为结构裂缝检测的重要步骤，存在裂缝目标识别与裂缝信息提取集成度不高的问题. 为此，提出一种基于计算机视觉与混合测量技术的结构裂缝识别方法. 首先，利用YOLOv8 （you only look once version 8）目标识别算法，实现结构裂缝的快速识别与定位；其次，基于稠密深度反向投影网络（D-DBPN）和UNet网络构建SR-UNet裂缝分割模型，并引入边界损失对原有损失函数进行改进，降低正负样本不平衡的影响，实现像素级裂缝提取；然后，结合连通域去噪、边缘检测等形态学技术，采用基于最短距离法与正交骨架法的混合方法对裂缝进行像素宽度测量；最后，利用LabelImg软件制作包含3 123张裂缝图像的识别定位数据集进行模型训练与测试. 研究结果表明：YOLOv8模型在裂缝测试集上的准确率为83.41%，召回率为84.93%，F₁分数为84%；裂缝像素宽度混合测量方法的拟裂缝宽度测量结果与人工识别方法基本一致，相对误差低于7.1%；本文方法能够实现裂缝识别定位、裂缝分割和像素宽度测量的一体化处理，对桥梁裂缝检测的发展具有较强的研究价值与应用前景.
- 裂缝识别 /
- 裂缝测量 /
- YOLOv8 /
- UNet /
- 超分辨图像
Abstract:
Concrete surface crack detection provides essential technical data and decision-making elements for the operation and maintenance of bridge structures. Crack identification is a key step in structural crack detection. However, the integration between crack target identification and information extraction is low. To this end, a new method for identifying structural cracks by using computer vision and hybrid measurement technology was proposed. Firstly, the You Only Look Once version 8 (YOLOv8) target recognition algorithm was employed to achieve rapid identification and localization of structural cracks. A super-resolution U-net (SR-UNet) crack segmentation model was developed based on the dense deep back-projection network (D-DBPN) and UNet, and boundary loss was introduced to improve the previous loss function, which addressed the imbalance between positive and negative samples and enabled precise pixel-level crack extraction. By using morphological techniques such as connected domain denoising and edge detection and a hybrid method of the shortest distance and orthogonal skeleton, the crack width at the pixel level was measured. A dataset of recognition and localization containing 3 123 crack images was created by using LabelImg software for model training and testing. The research results indicate that the YOLOv8 model achieves an accuracy of 83.41%, a recall rate of 84.93%, and an F₁ score of 84% on the crack test dataset. The simulated crack width measured by the hybrid crack width measurement method is consistent with that of manual identification, with a relative error of less than 7.1%. This method can integrate crack identification and localization, crack segmentation, and pixel width measurement, demonstrating significant research value and application potential for the development of bridge crack detection.
- crack recognition /
- crack measurement /
- YOLOv8 /
- UNet /
- super-resolution image

HTML全文

图 1 桥梁结构裂缝识别方法流程

Figure 1. Process of bridge structural crack identification

下载: 全尺寸图片幻灯片

图 2 YOLOv8模型

Figure 2. YOLOv8 model

下载: 全尺寸图片幻灯片

图 3 SR-UNet模型图

Figure 3. SR-UNet model

下载: 全尺寸图片幻灯片

图 4 损失函数曲线与mAP0.5曲线

Figure 4. Loss function curves and mAP0.5 curves

下载: 全尺寸图片幻灯片

图 5 不同光照强度下裂缝识别结果

Figure 5. Crack identification results under different light intensities

下载: 全尺寸图片幻灯片

图 6 SR图像与原图像的峰值信噪比与结构相似性指数

Figure 6. Peak signal-to-noise ratio and structural similarity index between SR image and original image

下载: 全尺寸图片幻灯片

图 7 图像分割结果的损失函数与IoU

Figure 7. Loss function and IoU for image segmentation results

下载: 全尺寸图片幻灯片

图 8 拟裂缝图像宽度测量

Figure 8. Width measurement of simulated crack image

下载: 全尺寸图片幻灯片

图 9 裂缝宽度测量方法比较

Figure 9. Comparison of crack width measurement methods

下载: 全尺寸图片幻灯片

图 10 裂缝检测全过程

Figure 10. Whole process of crack detection

下载: 全尺寸图片幻灯片

表 1 混淆矩阵

Table 1. Confusion matrix

预测值	真实值
	裂缝	背景
裂缝	TP	FP
背景	FN	TN

下载: 导出CSV

表 2 YOLOv5与YOLOv8模型比较

Table 2. Comparison of YOLOv5 and YOLOv8 models %

模型	P	R	F₁ 分数
YOLOv5-s	81.41	67.01	74
YOLOv8-s	81.69	84.02	83
YOLOv8-m	83.41	84.93	84

下载: 导出CSV

表 3 自动测量与人工测量对比

Table 3. Comparison of automatic and manual measurement

编号	像素数/个	测量宽度/mm		误差相对值/%
编号	像素数/个	自动测量	人工测量	误差相对值/%
1	30.0	18.8	18.6	0.81
2	49.0	30.6	29.6	3.46
3	28.5	17.8	16.8	6.03
4	30.5	19.1	17.8	7.09
5	30.0	18.8	19.4	3.35
6	36.0	22.5	21.6	4.17

下载: 导出CSV

参考文献(27)

[1]	李胜伟, 魏茂彬. 关于混凝土结构裂缝成因及预防措施研究[J]. 建筑结构, 2023, 53(增1): 1608-1611. LI Shengwei, WEI Maobin. Study on causes and prevention measures of concrete structure cracks[J]. Building Structure, 2023, 53(S1): 1608-1611.
[2]	覃潇, 申爱琴, 郭寅川, 等. 多场耦合下路面混凝土细观裂缝的演化规律[J]. 华南理工大学学报(自然科学版), 2017, 45(6): 81-88, 102. QIN Xiao, SHEN Aiqin, GUO Yinchuan, et al. Evolution rule of microcosmic cracks in pavement concrete under multi-field coupling[J]. Journal of South China University of Technology (Natural Science Edition), 2017, 45(6): 81-88, 102.
[3]	NAKAO M, HASEGAWA E, KUDO T, et al. Development of a bridge inspection support robot system using two-wheeled multicopters[J]. Journal of Robotics and Mechatronics, 2019, 31(6): 837-844. doi: 10.20965/jrm.2019.p0837
[4]	奚韵哲. 桥用爬壁机器人检测高墩裂缝技术研究[D]. 重庆: 重庆交通大学, 2019.
[5]	LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot MultiBox detector[C]//Computer Vision—ECCV 2016: 14th European Conference, Amsterdam, The Netherlands. Cham: Springer International Publishing, 2016: 21-37.
[6]	REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Las Vegas: IEEE, 2016: 779-788.
[7]	REDMON J, FARHADI A. YOLO9000: better, faster, stronger[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Honolulu: IEEE, 2017: 6517-6525.
[8]	GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation[C]//2014 IEEE Conference on Computer Vision and Pattern Recognition. Columbus: IEEE, 2014: 580-587.
[9]	GIRSHICK R. Fast R-CNN[C]//2015 IEEE International Conference on Computer Vision (ICCV). Santiago: IEEE, 2015: 1440-1448.
[10]	REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149. doi: 10.1109/TPAMI.2016.2577031
[11]	崔弥达, 王超, 陈金桥, 等. 基于ROS及YOLOv3的混凝土桥梁裂缝实时检测系统[J]. 东南大学学报(自然科学版), 2023, 53(1): 61-66. CUI Midai, WANG Chao, CHEN Jinqiao, et al. Real-time concrete bridge cracks detection system based on ROS and YOLOv3[J]. Journal of Southeast University (Natural Science Edition), 2023, 53(1): 61-66.
[12]	余加勇, 刘宝麟, 尹东, 等. 基于YOLOv5和U-Net3+的桥梁裂缝智能识别与测量[J]. 湖南大学学报(自然科学版), 2023, 50(5): 65-73. YU Jiayong, LIU Baolin, YIN Dong, et al. Intelligent identification and measurement of bridge cracks based on YOLOv5 and U-Net3+[J]. Journal of Hunan University (Natural Sciences), 2023, 50(5): 65-73.
[13]	LIU Y, ZHOU T, XU J Y, et al. Rotating target detection method of concrete bridge crack based on YOLO v5[J]. Applied Sciences, 2023, 13(20): 11118.1-11118.14.
[14]	RONNEBERGER O, FISCHER P, BROX T. U-net: convolutional networks for biomedical image segmentation[C]//Medical Image Computing and Computer-Assisted Intervention–MICCAI 2015: 18th International Conference. Munich: Springer, 2015: 234-241.
[15]	FU H X, MENG D, LI W H, et al. Bridge crack semantic segmentation based on improved Deeplabv3+ [J]. Journal of Marine Science and Engineering, 2021, 9(6): 671.1-671.14.
[16]	SU H F, WANG X, HAN T, et al. Research on a U-net bridge crack identification and feature-calculation methods based on a CBAM attention mechanism[J]. Buildings, 2022, 12(10): 1561.1-1561.18.
[17]	梁栋, 张伟, 于洋. 基于MU-net的混凝土桥裂缝识别方法[J]. 北京交通大学学报, 2022, 46(4): 105-112. doi: 10.11860/j.issn.1673-0291.20210081 LIANG Dong, ZHANG Wei, YU Yang. Crack identification method of concrete bridge based on MU-net[J]. Journal of Beijing Jiaotong University, 2022, 46(4): 105-112. doi: 10.11860/j.issn.1673-0291.20210081
[18]	TANG Y G, LIU C L, ZHANG X G. Single image super-resolution using Wasserstein generative adversarial network with gradient penalty[J]. Pattern Recognition Letters, 2022, 163: 32-39. doi: 10.1016/j.patrec.2022.09.012
[19]	HARIS M, SHAKHNAROVICH G, UKITA N. Deep back-projection networks for super-resolution[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Salt Lake City: IEEE, 2018: 1664-1673.
[20]	KONDO Y, UKITA N. Crack segmentation for low-resolution images using joint learning with super- resolution[C]//2021 17th International Conference on Machine Vision and Applications (MVA). Aichi: IEEE, 2021: 1-6.
[21]	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. 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.
[22]	TERVEN J, CÓRDOVA-ESPARZA D M, ROMERO-GONZÁLEZ J A. A comprehensive review of yolo architectures in computer vision: From yolov1 to yolov8 and yolo-nas[J]. Machine Learning and Knowledge Extraction, 2023, 5(4): 1680-1716. doi: 10.3390/make5040083
[23]	ONG J C H, ISMADI M P, WANG X. A hybrid method for pavement crack width measurement[J]. Measurement, 2022, 197: 111260.1-111260.16.
[24]	LIU C, TANG C S, SHI B, et al. Automatic quantification of crack patterns by image processing[J]. Computers & Geosciences, 2013, 57: 77-80.
[25]	QIU S, WANG W J, WANG S F, et al. Methodology for accurate AASHTO PP67-10—based cracking quantification using 1-mm 3D pavement images[J]. Journal of Computing in Civil Engineering, 2017, 31(2): 04016056.1-04016056.9.
[26]	KERVADEC H, BOUCHTIBA J, DESROSIERS C, et al. Boundary loss for highly unbalanced segmentation[J]. Medical Image Analysis, 2021, 67: 101851.1-101851.11.
[27]	崔晓宁, 王起才, 李盛, 等. 基于YOLO-v5的双块式轨枕裂缝智能识别[J]. 铁道学报, 2022, 44(4): 104-111. CUI Xiaoning, WANG Qicai, LI Sheng, et al. Intelligent recognition of cracks in double block sleeper based on YOLO-v5[J]. Journal of the China Railway Society, 2022, 44(4): 104-111.