An advanced method for surface damage detection of concrete structures in low-light environments based on image enhancement and object detection networks

Abstract

Surface damage detection in concrete structures is critical for maintaining structural integrity, yet current object detection algorithms often struggle in low-light environments. To address this challenge, this study proposed a methodology that integrates image enhancement and object detection networks to improve damage identification in such conditions. Specifically, we employ the self-calibrated illumination (SCI) model to reconstruct low-light images, which are then processed by an improved YOLOv5-based network, YOLOv5-GAM-ASFF, incorporating a global attention mechanism (GAM) and adaptive spatial feature fusion (ASFF). The performance of YOLOv5-GAM-ASFF is evaluated on a dataset of concrete structure damage images, demonstrating its superiority over YOLOv5s, YOLOv6s, and YOLOv7-tiny. The results show that YOLOv5-GAM-ASFF achieves a mAP@0.5 of 79.1%, surpassing the other models by 1.3%, 3.3%, and 5.8%, respectively. This approach provides a reliable solution for surface damage detection in low-light environments, advancing the field of structural health monitoring by improving detection accuracy under challenging conditions.

References

[1] Ackar H, Almisreb AA, Saleh MA (2019) A review on image enhancement techniques. Southeast Eur J Soft Comput. https://api.semanticscholar.org/CorpusID:149589363

[2] Agustsson E, Timofte R (2017) Ntire 2017 challenge on single image super-resolution: dataset and study. (2017 IEEE conference on computer vision and pattern recognition workshops, Honolulu, USA). https://doi.org/10.1109/CVPRW.2017.150

[3] Chao X, Wei W, Lu D. Crack detection algorithm for concrete structures based on super-resolution reconstruction and segmentation network. Autom Constr, 2022, 140(104346): 0926-5805

[4] Dadboud F, Patel V, Mehta V. Single-stage UAV detection and classification with YOLOv5: mosaic data augmentation and Panet. In: 2021 17th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS), 2021 1-8

[5] Dhal KG, Ray S, Das A (2019) A survey on nature-inspired optimization algorithms and their application in image enhancement domain. Springer Netherlands 26:1607–1638. https://doi.org/10.1007/s11831-018-9289-9

[6] He ZL, Jiang S, Zhang J. Automatic damage detection using anchor-free method and unmanned surface vessel. Autom Constr, 2022, 133: 104017

[7] Hoffmann N, Liu Y, Shao Z (2021) Global attention mechanism: retain information to enhance channel-spatial interactions. arxiv preprint arxiv:2112.05561. https://doi.org/10.48550/arXiv.2112.05561

[8] Hofinger P, Klemmt HJ, Ecke S. Application of YOLOv5 for point label based object detection of black pine trees with vitality losses in UAV data. Remote Sens, 2023, 15(8): 1964

[9] Hoshyar AN, Yu Y, Samali B, Zhang G. Corrosion and coating defect assessment of coal handling and preparation plants (CHPP) using an ensemble of deep convolutional neural networks and decision-level data fusion. Neural Comput Appl, 2023, 35(25): 18697-18718

[10] Hu CJ, Jiang TY, Li LY (2024) Complex background segmentation for noncontact cable vibration frequency estimation using semantic segmentation and complexity pursuit algorithm. J Civ Struct Health Monit. https://doi.org/10.1007/s13349-024-00798-6

[11] Jiang TY, Li LY, Samali B, Yu Y (2024) Lightweight object detection network for multi-damage recognition of concrete bridges in complex environments. Comput Aided Civ Infrastruct Eng. https://doi.org/10.1111/mice.13219

[12] Kim D, Park S, Kang D. Improved center and scale prediction-based pedestrian detection using convolutional block. In: 2019 IEEE 9th International Conference on Consumer Electronics (ICCE-Berlin), 2019 https://ieeexplore.ieee.org/document/8966154

[13] Kingma DP, Ba J (2014) Adam: a method for stochastic optimization. arxiv preprint arxiv:1412.6980. https://doi.org/10.48550/arXiv.1412.6980

[14] Luo ZX, Ma TY, Ma L. PIA: parallel architecture with illumination allocator for joint enhancement and detection in low-light. Assoc Comput Mach, 2022, 9: 2070-2078

[15] Ma L, Ma T, Liu R (2022) Toward fast, flexible, and robust low-light image enhancement. arxiv preprint arxiv:2204.10137. https://doi.org/10.48550/arXiv.2204.10137

[16] Ni FT, Zhang J, Chen ZQ. Zernike-moment measurement of thin-crack width in images enabled by dual-scale deep learning. Comput Aided Civ Infrastruct Eng, 2018, 34: 367-384

[17] Ooi YK, Ibrahim H. Deep learning algorithms for single image super-resolution: a systematic review. Electronics, 2021, 10(7): 867

[18] Pan Z, DL L. EPSA-YOLO-V5s: A novel method for detecting the survival rate of rapeseed in a plant factory based on multiple guarantee mechanisms. Comput Electron Agric, 2022, 193: 106714

[19] Park J, Woo S, Lee JY (2018) BAM: bottleneck attention module. arxiv preprint arxiv: 1807.06514. https://doi.org/10.48550/arXiv.1807.06514

[20] Peng YP, Wang WJ, Tang Z. Non-uniform illumination image enhancement for surface damage detection of wind turbine blades. Mech Syst Signal Process, 2022, 170: 108797

[21] Qi YL, Yang Z, Sun WH. A comprehensive overview of image enhancement techniques. Arch Comput Methods Eng, 2022, 29: 583-607

[22] Qin S, Qi T, Lei B. Rapid and automatic image acquisition system for structural surface defects of high-speed rail tunnels. KSCE J Civ Eng, 2024, 28(2): 967-989

[23] Qiu ML, Huang L, Tang BH. ASFF-YOLOv5: Multielement detection method for road traffic in UAV images based on multiscale feature fusion. Remote Sens, 2022, 14(14): 3498

[24] Ren S, He KM, Zhang XY. Spatial pyramid pooling in deep convolutional networks for visual recognition. IEEE Trans Pattern Anal Mach Intell, 2015, 37(9): 1904-1916

[25] Sathya K, Sangavi D, Sridharshini P. Improved image-based super resolution and concrete crack prediction using pre-trained deep learning models. J Soft Comput Civ Eng, 2022, 4(3): 40-51 https://api.semanticscholar.org/CorpusID:225525662

[26] Shen L, Hu J, Sun G. Squeeze-and-excitation networks. In: Proceedings of the IEEE conference on computer vision and pattern recognition 2018, 2018 7132-7141

[27] Shu ZT, Zhang ZB, Song YZ. Low-light image object detection based on improved YOLOv5 algorithm. Laser Optoelectron Prog (in Chinese), 2023, 60(4): 77-84 https://kns.cnki.net/kcms/detail/31.1690.TN.20220713.1846.489.html

[28] Silva WRLD, Lucena DSD. Concrete cracks detection based on deep learning image classification. Proceedings, 2018, 2(8): 489

[29] Wang WH, Xie EZ, Song XG. Efficient and accurate arbitrary-shaped text detection with pixel aggregation network. In: IEEE/CVF international conference on computer vision, 2019

[30] Woo S, Park J, Lee JY. Cbam: convolutional block attention module. In: European conference on computer vision (ECCV), 2018

[31] Wu W, Weng J, Zhang P. “URetinex-Net: Retinex-based Deep Unfolding Network for Low-light Image Enhancement”. In: 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2022 5891-5900

[32] Yang GH, Feng W, Jin JT. Face mask recognition system with YOLOv5 based on image recognition. In: IEEE 6th International Conference on Computer and Communications (ICCC), 2020 https://api.semanticscholar.org/CorpusID:231919807

[33] Yu Y, Samali B, Rashidi M. Vision-based concrete crack detection using a hybrid framework considering noise effect. J Build Eng, 2022, 61: 105246

[34] Zhang CW, Yu Y, Yousefi AM. Compressive strength evaluation of cement-based materials in sulphate environment using optimized deep learning technology. Dev Built Environ, 2023, 16: 100298

[35] Zhao ZQ, Zheng P, Xu ST. Object detection with deep learning: a review. IEEE Trans Neural Netw Learn Syst, 2019, 30(11): 3212-3232

RIGHTS & PERMISSIONS

The Author(s)