Page 124 - 2025年第56卷第10期

［ 22］ LI Y，BAO T，HUANG X，et al. Underwater crack pixel-wise identification and quantification for dams via light⁃
weight semantic segmentation and transfer learning［J］ Automation in Construction，2022，144：104600.
.
［ 23］ 李扬涛，赵海涛，魏洋，等 . 视觉驱动的涉水混凝土结构水下多类别缺陷辨识和区域弱监督定位方法［J］ 应用
.
基础与工程科学学报，2025，33（1）：60-75.
［ 24］ LI S， ZHAO X. High-resolution concrete damage image synthesis using conditional generative adversarial network
［J］ Automation in Construction，2023，147：104739.
.
.
［ 25］ ZAVRTANIK V，KRISTAN M，SKOCAJ D. Reconstruction by inpainting for visual anomaly detection［J］ Pattern
Recognition，2021，112：107706.
.
［ 26］ 王立成，武少赟 . 混凝土劈拉开裂和裂缝自愈合机理［J］ 水利学报，2019，50（7）：787-797.
.
［ 27］ DUNG C，ANH L. Autonomous concrete crack detection using deep fully convolutional neural network［J］ Automa⁃
tion in Construction，2019，99：52-58.
［ 28］ PAYAB M， ABBASINA R， KHANZADI M. A brief review and a new graph-based image analysis for concrete
.
crack quantification［J］ Archives of Computational Methods in Engineering，2019，26（2）：347-365.
［ 29］ IYER S， SINHA S. A robust approach for automatic detection and segmentation of cracks in underground pipeline
images［J］ Image and Vision Computing，2005，23（10）：921-933.
.
.
［ 30］ HARALICK R，STERNBERG S，ZHUANG X. Image-analysis using mathematical morphology［J］ IEEE Transac⁃
tions on Pattern Analysis and Machine Intelligence，1987，9（4）：532-550.
.
［ 31］ OTSU N. Threshold selection method from gray-level histograms［J］ IEEE Transactions on Systems Man and Cyber⁃
netics，1979，9（1）：62-66.
［ 32］ TANG H， XU D， SEBEL N， et al. Attention-guided generative adversarial networks for unsupervised image-to-
.
image translation［C］/2019 International Joint Conference on Neural Networks （IJCNN） 2019.
/
/
［ 33］ ISOLA P， ZHU J， ZHOU T， et al. Image-to-image translation with conditional adversarial networks［C］/30th
.
IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR） 2017.
［ 34］ GUO L，LI R，JIANG B. A cascade broad neural network for concrete structural crack damage automated classifica⁃
tion［J］ IEEE Transactions on Industrial Informatics，2021，17（4）：2737-2742.
.
An unsupervised crack segmentation method based on
attention-guided reconstruction network

1 1 2 3

CHEN Dong ，KANG Fei ，LI Xinyu ，ZHANG Yaping

（1. School of Infrastructure Engineering，Dalian University of Technology，Dalian 116024，China；

2. China Yangtze Power Co.，Ltd.，Yichang 443000，China；3. China Three Gorges Corporation，Beijing 100038，China）

Abstract：To address issues such as sample scarcity，annotation difficulty，and background noise interference in
dam surface crack detection，an unsupervised crack segmentation method based on an attention-guided reconstruc⁃

tion network，named AGR-CrackNet，is proposed. During the training phase，masks conforming to crack morphology

characteristics are randomly generated to create damaged process of intact concrete images，thereby creating training

samples for the AGR-CrackNet. During testing，the pre-trained model reconstructs crack images，and precise seg⁃

mentation is achieved by calculating the absolute differences between the reconstructed and original images，com⁃
bined with the Otsu algorithm. Ablation experiments demonstrate that the proposed mask generation process and train⁃
ing mechanism significantly improve segmentation accuracy while effectively suppressing noise-induced errors. AGR-
CrackNet improves F scores by ranges of 10.0%-66.8% on dam surface and 26.4%-31.2% on underwater datasets，
1

compared to other unsupervised methods and surpassing most supervised algorithms. Furthermore，analysis indicates
that the high-quality labeled data generated by AGR-CrackNet can be utilized to train supervised networks，further

expanding its application scenarios.

Keywords：dam surface crack；unsupervised crack segmentation；reconstruction network；attention-guided mecha⁃

nism；defect detection
（责任编辑：韩 昆）
— 1371 —