Underwater bridge pier crack image generation based on improved CycleGAN

doi:10.3785/j.issn.1008-973X.2025.09.008

Journal of ZheJiang University (Engineering Science)

2025, Vol. 59

Issue (9): 1846-1855 DOI: 10.3785/j.issn.1008-973X.2025.09.008

Underwater bridge pier crack image generation based on improved CycleGAN

Zhenming LV1(

),Shaojiang DONG1,*(

),Jingyao HE2,Jinlong YANG3,Jiawei ZHANG1

1. School of Mechatronics and Vehicle Engineering, Chongqing Jiaotong University, Chongqing 400074, China
2. School of River and Ocean Engineering, Chongqing Jiaotong University, Chongqing 400074, China
3. School of Traffic and Transportation, Chongqing Jiaotong University, Chongqing 400074, China

Download:

HTML

PDF(3634KB) HTML
Export: BibTeX | EndNote (RIS)

Abstract

Deep learning-based underwater bridge pier crack detection relies heavily on a large number of crack images. However, pier cracks are scarce and disturbed by the water environment, resulting in poor crack image quality. Thus, an improved CycleGAN network was proposed to generate underwater bridge pier crack image data. Underwater bridge pier wall image data were collected by an underwater robot and combined with bridge crack data to form a dataset. The scSE attention was added to the generator of CycleGAN, and the DehazeFormer module was added at the neck, to ensure the quality of the generated data. These enhancements improved the quality of the generated underwater bridge pier crack images, enabling better distribution and discriminative ability in the feature space. A pixel-aware discriminator was used to accurately discriminate the generated images. When compared with mainstream image conversion algorithms, the proposed method demonstrated superiority in terms of underwater image quality (UIQM), underwater color quality (UCIQE), and peak signal-to-noise ratio (PSNR). The UIQM score reached 0.818, the UCIQE score reached 0.443, and the PSNR metric score reached 24.673. To verify the crack image quality, target detection tasks were performed using both the generated underwater images and real underwater bridge pier crack data collected by the robot. The results showed that the F1 score and mAP50 indicator scores differed by less than 0.1%. The proposed method was expected to solve the problem of insufficient data for target detection tasks and provide strong data support for the safety assessment of underwater bridge piers.

Key words： underwater bridge pier crack data scarcity CycleGAN scSE attention DehazeFormer module pixel-aware discriminator

Received: 26 September 2024 Published: 25 August 2025

CLC:	U 446
	TP 391.41

Fund: 重庆市自然科学基金创新发展联合基金资助项目(CSTB2024NSCQ-LZX0024)；重庆市教育委员会科学技术研究资助项目(KJZD-K202300711)；重庆市研究生科研创新资助项目(CYS240489).

Corresponding Authors: Shaojiang DONG E-mail: 13946501711@163.com;dongshaojiang100@163.com

	Service
	E-mail this article
	Add to my bookshelf
	Add to citation manager
	E-mail Alert
	RSS
	Articles by authors
	Zhenming LV
	Shaojiang DONG
	Jingyao HE
	Jinlong YANG
	Jiawei ZHANG

Cite this article:

Zhenming LV,Shaojiang DONG,Jingyao HE,Jinlong YANG,Jiawei ZHANG. Underwater bridge pier crack image generation based on improved CycleGAN. Journal of ZheJiang University (Engineering Science), 2025, 59(9): 1846-1855.

URL:

https://www.zjujournals.com/eng/10.3785/j.issn.1008-973X.2025.09.008 OR https://www.zjujournals.com/eng/Y2025/V59/I9/1846

基于改进CycleGAN的水下桥墩裂缝图像生成

基于深度学习的水下桥墩裂缝检测依赖大量的裂缝图像，然而桥墩裂缝稀缺且受水体环境干扰导致裂缝图像质量不佳，为此提出改进的CycleGAN网络用于生成水下桥墩裂缝图像数据. 通过水下机器人采集水下桥墩壁面图像数据，结合桥梁裂缝数据构建数据集. 为了保证生成数据质量，在CycleGAN的生成器中添加scSE注意力，并在颈部处添加DehazeFormer模块，使生成的水下桥墩裂缝的图像质量提升，在特征空间中具有更好的分布和区分度. 采用像素感知判别器对生成的图像进行精准判别. 采用所提方法，水下图像质量UIQM、水下色彩质量UCIQE和峰值信噪比PSNR分别达到0.818、0.443和24.673，生成的水下桥墩裂缝效果优于其他主流图像转换算法的. 为了验证裂缝图像质量，结合机器人采集的水下桥墩裂缝数据，采用目标检测任务对比生成的水下图像质量和真实的水下桥墩裂缝质量，结果表明，F1分数和mAP50指标分数相差均小于0.1%. 所提方法有望解决目标检测任务数据不足问题，为水下桥墩的安全评估提供有力的数据支持.

关键词： 水下桥墩裂缝, 数据稀缺, CycleGAN, scSE注意力, DehazeFormer模块, 像素感知判别器

Fig.1 Network framework for crack generation in underwater bridge piers

Fig.2 CycleGAN network framework for generating cracks in underwater bridge piers

Fig.3 Underwater bridge pier image data acquisition equipment and data collection process

Fig.4 Dataset generated through improved CycleGAN and real underwater bridge pier crack data

Tab.1 Quantitative comparison of multiple algorithms for generating underwater bridge pier cracks

Fig.5 Visual comparison results of multiple algorithms

Tab.2 Results of unreferenced metric assessment of real underwater bridge pier cracks and pier cracks generated by proposed method

Tab.3 Results of ablation experiment indicators

Fig.6 Visual comparison of improved CycleGAN generator ablation experiments

Tab.4 Validation of target detection effect for generated and real samples

Fig.7 Visualization of target detection of generated underwater bridge pier cracks

Fig.8 Comparison of thermograms under task of detecting cracks in underwater bridge pier


[1]	LI X, MENG Q, WEI M, et al Identification of underwater structural bridge damage and BIM-based bridge damage management[J]. Applied Sciences, 2023, 13 (3): 1348 doi: 10.3390/app13031348

[2]	ORINAITĖ U, KARALIŪTĖ V, PAL M, et al Detecting underwater concrete cracks with machine learning: a clear vision of a murky problem[J]. Applied Sciences, 2023, 13 (12): 7335 doi: 10.3390/app13127335

[3]	KOV’ARI K, PETER G Continuous strain monitoring in the rock foundation of a large gravity dam[J]. Rock Mechanics and Rock Engineering, 1983, 16 (3): 157- 171 doi: 10.1007/BF01033277

[4]	FAN X, CAO P, SHI P, et al An underwater dam crack image segmentation method based on multi-level adversarial transfer learning[J]. Neurocomputing, 2022, 505: 19- 29 doi: 10.1016/j.neucom.2022.07.036

[5]	HUANG B, KANG F, LI X, et al Underwater dam crack image generation based on unsupervised image-to-image translation[J]. Automation in Construction, 2024, 163: 105430 doi: 10.1016/j.autcon.2024.105430

[6]	YE X, LUO K, WANG H, et al An advanced AI-based lightweight two-stage underwater structural damage detection model[J]. Advanced Engineering Informatics, 2024, 62: 102553 doi: 10.1016/j.aei.2024.102553

[7]	程风雯, 甘进, 李星, 等基于DCGAN的水下结构物表面缺陷图像生成[J]. 长江科学院院报, 2023, 40 (9): 155- 161 CHENG Fengwen, GAN Jin, LI Xing, et al Image generation for surface defects of underwater structures based on deep convolutional generative adversarial networks[J]. Journal of Changjiang River Scientific Research Institute, 2023, 40 (9): 155- 161 doi: 10.11988/ckyyb.20220421

[8]	王桂平, 陈旺桥, 杨建喜, 等基于迁移学习的桥梁表观病害检测技术研究[J]. 铁道科学与工程学报, 2022, 19 (6): 1638- 1646 WANG Guiping, CHEN Wangqiao, YANG Jianxi, et al A bridge surface distress detection technology based on transfer learning[J]. Journal of Railway Science and Engineering, 2022, 19 (6): 1638- 1646

[9]	ZHAO S, SHADABFAR M, ZHANG D, et al Deep learning-based classification and instance segmentation of leakage-area and scaling images of shield tunnel linings[J]. Structural Control and Health Monitoring, 2021, 28 (6): e2732

[10]	马金祥, 范新南, 吴志祥, 等暗通道先验的大坝水下裂缝图像增强算法[J]. 中国图象图形学报, 2016, 21 (12): 1574- 1584 MA Jinxiang, FAN Xinnan, WU Zhixiang, et al Underwater dam crack image enhancement algorithm based on improved dark channel prior[J]. Journal of Image and Graphics, 2016, 21 (12): 1574- 1584 doi: 10.11834/jig.20161202

[11]	XIN G, FAN X, SHI P, et al A fine extraction algorithm for image-based surface cracks in underwater dams[J]. Measurement Science and Technology, 2023, 34 (3): 035402 doi: 10.1088/1361-6501/ac9db2

[12]	QI Z, LIU D, ZHANG J, et al Micro-concrete crack detection of underwater structures based on convolutional neural network[J]. Machine Vision and Applications, 2022, 33 (5): 74 doi: 10.1007/s00138-022-01327-5

[13]	雍子叶, 郭继昌, 李重仪融入注意力机制的弱监督水下图像增强算法[J]. 浙江大学学报: 工学版, 2021, 55 (3): 555- 562,570 YONG Ziye, GUO Jichang, LI Chongyi Weakly supervised underwater image enhancement algorithm incorporating attention mechanism[J]. Journal of Zhejiang University: Engineering Science, 2021, 55 (3): 555- 562,570

[14]	范新南, 顾丽萍, 巫鹏, 等一种仿水下生物视觉的大坝裂缝图像增强算法[J]. 光电子激光, 2014, 25 (2): 372- 377 FAN Xinnan, GU Liping, WU Peng, et al A dam crack image enhancement algorithm based on underwater biological vision[J]. Journal of Optoelectronics Laser, 2014, 25 (2): 372- 377

[15]	温佩芝, 陈君谋, 肖雁南, 等基于生成式对抗网络和多级小波包卷积网络的水下图像增强算法[J]. 浙江大学学报: 工学版, 2022, 56 (2): 213- 224 WEN Peizhi, CHEN Junmou, XIAO Yannan, et al Underwater image enhancement algorithm based on GAN and multi-level wavelet CNN[J]. Journal of Zhejiang University: Engineering Science, 2022, 56 (2): 213- 224

[16]	ZHU J Y, PARK T, ISOLA P, et al. Unpaired image-to-image translation using cycle-consistent adversarial networks [C]//Proceedings of the IEEE International Conference on Computer Vision. Venice: IEEE, 2017: 2242–2251.

[17]	ROY A G, NAVAB N, WACHINGER C. Concurrent spatial and channel ‘squeeze&excitation’ in fully convolutional networks [C]// Medical Image Computing and Computer Assisted Intervention – MICCAI 2018. Cham: Springer, 2018: 421–429.

[18]	SONG Y, HE Z, QIAN H, et al Vision transformers for single image dehazing[J]. IEEE Transactions on Image Processing, 2023, 32: 1927- 1941 doi: 10.1109/TIP.2023.3256763

[19]	LIU Z, LIN Y T, CAO Y, et al. Swin transformer: hierarchical vision transformer using shifted windows [C]// Proceedings of the IEEE/CVF International Conference on Computer Vision. Montreal: IEEE, 2021: 10012-10022.

[20]	BA J L, KIROS J R, HINTON G E. Layer normalization [EB/OL]. [2024-06-01]. https://arxiv.org/abs/1607.06450.

[21]	HENDRYCKS D, GIMPEL K. Gaussian error linear units (gelus)[EB/OL]. [2024-06-01]. https://arxiv.org/abs/1606.08415v5.

[22]	WANG P, CHEN P, YUAN Y, et al. Understanding convolution for semantic segmentation [C]// IEEE Winter Conference on Applications of Computer Vision. Lake Tahoe: IEEE, 2018: 1451–1460.

[23]	ENGIN D, GENÇ A, KEMAL EKENEL H. Cycle-dehaze: Enhanced cyclegan for single image dehazing [C]// Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops. Salt Lake City: IEEE, 2018: 825-833.

[24]	WANG Y, YAN X, GUAN D, et al Cycle-SNSPGAN: towards real-world image dehazing via cycle spectral normalized soft likelihood estimation patch GAN[J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23 (11): 20368- 20382 doi: 10.1109/TITS.2022.3170328

[25]	WANG Y, YAN X, WANG F L, et al UCL-dehaze: toward real-world image dehazing via unsupervised contrastive learning[J]. IEEE Transactions on Image Processing, 2024, 33: 1361- 1374 doi: 10.1109/TIP.2024.3362153

[26]	YANG M, SOWMYA A An underwater color image quality evaluation metric[J]. IEEE Transactions on Image Processing, 2015, 24 (12): 6062- 6071 doi: 10.1109/TIP.2015.2491020

[27]	PANETTA K, GAO C, AGAIAN S Human-visual-system-inspired underwater image quality measures[J]. IEEE Journal of Oceanic Engineering, 2016, 41 (3): 541- 551 doi: 10.1109/JOE.2015.2469915

[28]	YANG X, LI H, YU Y, et al Automatic pixel-level crack detection and measurement using fully convolutional network[J]. Computer-Aided Civil and Infrastructure Engineering, 2018, 33 (12): 1090- 1109 doi: 10.1111/mice.12412

[29]	WANG W, YANG Y A color image fusion model by saturation-value total variation[J]. Journal of Computational and Applied Mathematics, 2024, 446: 115832 doi: 10.1016/j.cam.2024.115832

[30]	ZHAO Y, LV W, XU S, et al. Detrs beat yolos on real-time object detection [C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seattle: IEEE, 2024: 16965-16974.

[1]	Zhenming LV,Shaojiang DONG,Zongyou XIA,Xiaoyan MOU,Mingquan WANG. Multi-distortion type underwater image enhancement based on improved CycleGAN[J]. Journal of ZheJiang University (Engineering Science), 2025, 59(6): 1148-1158.

Viewed

Full text

Abstract

Cited

Shared

Discussed