在自然环境中，金属涂层材料可能会因划痕导致周边剥落或腐蚀，直接影响金属的使用寿命，所以提前对金属涂层剥落或腐蚀进行检测与评估具有非常重要的意义。目前已存在的检测与评估方法较为繁琐、耗时，很难快速得到较为准确的金属涂层剥落与腐蚀的等级信息。针对以上问题，本文提出了一种基于图像处理的金属涂层剥落与腐蚀等级评定方法，实现了对金属涂层剥落与腐蚀严重程度的量化与评估。在实际工程应用中，该研究具有重要的实用价值和广阔的应用前景。

针对金属涂层缺陷分割中存在特征提取能力弱和分割精度低的问题，提出了一种改进的U²-Net分割模型。首先在U型残差块（RSU）中嵌入改进的增大感受野模块（Receptive Field Block light，RFB_l），组成新的特征提取层——RFB_l-RSU，增强对细节特征的学习能力，解决了网络由于感受野受限造成分割精度低的问题；其次在U²-Net分割模型的解码阶段引入有效的边缘增强注意力机制（Contour Enhanced Attention，CEA）抑制网络中的冗余特征，获取具有详细位置信息的特征注意力图，增强了边界与背景信息的差异性，从而达到更精确的分割效果；最后，将所提方法与常用分割网络对比分析。实验结果表明，该模型在三个金属涂层剥落与腐蚀数据集上的平均交并比、准确率、查准率、召回率和F₁-measure分别达到80.33%、96.34%、87.29%、84.44%和85.85%，相比于常用的FCN、SegNet、U-Net以及U²-Net分割网络的性能都有所提高。

针对目前金属涂层剥落与腐蚀的宽度计算方法存在误差较大、效率低的问题，提出一种遍历取点计算金属涂层剥落与腐蚀区域宽度的方法。主要利用图像处理的方法沿金属涂层剥落与腐蚀区域进行遍历，均匀选取7个计算点，计算其对应的平均宽度，最后根据国际标准ISO4628-8对金属涂层剥落与腐蚀严重程度进行等级评定。实验表明，该方法相比于基于划线长度与面积关系方法和基于最大内接圆方法，计算的宽度误差最低，最接近真实值。并且与专家等级评定结果进行比较，本文方法不仅一致性好，评级平均正确率达到95%，效率更高，节省成本，能够有效满足工程应用要求。

In natural environments, metal coating materials may suffer from scratches that can lead to nearby peeling or corrosion, directly affecting the service life of the metal. Therefore, it is important to detect and assess the peeling and corrosion of metal coatings in advance. However, existing detection and assessment methods are cumbersome and time-consuming, making it difficult to obtain accurate information about the severity of metal coating peeling and corrosion quickly. To address this issue, this paper proposes a method for evaluating the level of metal coating peeling and corrosion based on image processing, which enables quantification and assessment of the severity of peeling and corrosion. In practical engineering applications, this study has important practical value and broad application prospects.

To address the problems of weak feature extraction and low segmentation accuracy in metal coating defect segmentation, an improved U²-Net segmentation model is proposed. First, an improved Receptive Field Block light (RFB_l) module is embedded in the U-shaped residual block (RSU) to form a new feature extraction layer, RFB_l-RSU, which enhances the network's ability to learn detail features and solves the problem of low segmentation accuracy caused by limited receptive fields. Second, an effective Contour Enhanced Attention (CEA) mechanism is introduced in the decoding stage of the U²-Net segmentation model to suppress redundant features in the network, obtain feature attention maps with detailed position information, and enhance the difference between boundary and background information, thereby achieving more accurate segmentation results. Finally, the proposed method is compared with commonly used segmentation networks. Experimental results show that the proposed model achieves average Intersection over Union (IoU), accuracy, precision, recall, and F₁-measure of 80.33%, 96.34%, 87.29%, 84.44%, and 85.85%, respectively, on three metal coating peeling and corrosion datasets, outperforming commonly used FCN, SegNet, U-Net, and U²-Net segmentation networks.

To address the problem of large errors and low efficiency in current metal coating peeling and corrosion width calculation methods, a traversal point selection method is proposed to calculate the width of metal coating peeling and corrosion regions. This method uses image processing to traverse along the metal coating peeling and corrosion regions, uniformly select seven calculation points, calculate their corresponding average widths, and finally evaluate the severity of metal coating peeling and corrosion according to international standard ISO4628-8. Experimental results show that compared to methods based on line length and area relationships and maximum inscribed circle methods, this method has the lowest width calculation error and is closest to the true value. Furthermore, compared with expert rating results, the proposed method has good consistency and an average rating accuracy of 95%, making it more efficient and cost-effective for practical engineering applications.

[1] Hanawa T. Overview of metals and applications[M]// Metals for biomedical devices. Woodhead Publishing, 2019: 3-29.

[2] Hou X, Gao L, Cui Z, et al. Corrosion and protection of metal in the seawater desalination[C]// IOP conference series: earth and environmental science. IOP Publishing, 2018, 108: 022-037.

[3] Harsimran S, Santosh K, Rakesh K. Overview of corrosion and its control: A critical review[J]. Proc. Eng. Sci, 2021, 3(1): 13-24.

[4] Whitfield M J, Bono D, Wei L, et al. High-throughput corrosion quantification in varied microenvironments[J]. Corrosion science, 2014, 88(54): 481-486.

[5] Blanchard P J, Hill D J, Bretz G T, et al. Evaluation of corrosion protection methods for magnesium alloys in automotive applications[J]. Essential Readings in Magnesium Technology, 2016: 593-598.

[6] Maples A L, Williams E B, Rawlins J W. Understanding scribe profile and tool type effects on visual corrosion assessments[J]. Journal of Failure Analysis and Prevention, 2018, 18(6): 1401-1410.

[7] Liao K W, Lee Y T. Detection of rust defects on steel bridge coatings via digital image recognition[J]. Automation in Construction, 2016, 71(2): 294-306.

[8] 龚应忠, 李子存, 冯新泸, 等．基于颜色特征的铜片腐蚀结果评价[J]．腐蚀与防护, 2013, 34(2): 117-120+150．

[9] Bondada V, Pratihar D K, Kumar C S. Detection and quantitative assessment of corrosion on pipelines through image analysis[J]. Procedia Computer Science, 2018, 133(10): 804-811.

[10] Dolezel P, Rozsival P, Rozsivalova V, et al. Automatized approach to assessment of degree of delamination around a scribe[C]// Intelligent Systems and Applications: Proceedings of the 2018 Intelligent Systems Conference (IntelliSys) Volume 1. Springer International Publishing, 2019: 227-236.

[11] Ahuja S K, Shukla M K, Ravulakollu K K. Neural Network-Based Surface Corrosion Classification on Metal Articles[C]// Proceedings of International Conference on Computational Intelligence, Data Science and Cloud Computing: IEM-ICDC 2020. Springer Singapore, 2021: 115-124.

[12] Sanchez G, Aperador W, Cerón A. Corrosion grade classification: a machine learning approach[J]. Indian Chemical Engineer, 2020, 62(3): 277-286.

[13] 姜卫. 基于改进型YOLOv3和模型轻量化的锈蚀检测及应用[D]. 南京: 南京邮电大学, 2021.

[14] 石林. 基于深度学习的金属表面腐蚀等级检测研究[D]. 重庆: 重庆理工大学, 2021.

[15] Rozsivalova V, Dolezel P, Stursa D. et al. Sequence of U-Shaped Convolutional Networks for Assessment of Degree of Delamination Around Scribe[J]. Int J Comput Intell Syst, 2022, 15(1): 76.

[16] Xu H, Xu X, Zuo Y. Applying morphology to improve Canny operator's image segmentation method[J]. The Journal of Engineering, 2019, 2019(23): 8816-8819.

[17] Liang Y, Xu K, Zhou P. Mask gradient response-based threshold segmentation for surface defect detection of milled aluminum ingot[J]. Sensors, 2020, 20(16): 4519.

[18] Vorobel R, Ivasenko I, Berehulyak O, et al. Segmentation of rust defects on painted steel surfaces by intelligent image analysis[J]. Automation in Construction, 2021, 123(30): 103515.

[19] 朱昱, 潘耀忠, 张杜娟. 基于深度卷积网络和分水岭分割的耕地地块识别方法[J]. 地球信息科学学报, 2022, 24(12): 2389-2403.

[20] Long J, Shelhamer E, Darrell T. Fully convolutional networks for semantic segmentation[C]// Proceedings of the IEEE conference on computer vision and pattern recognition. 2015: 3431-3440.

[21] 王达磊，彭博，潘玥，等．基于深度神经网络的锈蚀图像分割与定量分析[J]．华南理工大学学报（自然科学版），2018, 46(12): 121-127.

[22] Liu L, Tan E, Zhen Y, et al. AI-facilitated coating corrosion assessment system for productivity enhancement[C]// 2018 13th IEEE Conference on Industrial Electronics and Applications (ICIEA). IEEE, 2018: 606-610.

[23] Huang Z, Wang X, Huang L, et al. Ccnet: Criss-cross attention for semantic segmentation[C]// Proceedings of the IEEE/CVF international conference on computer vision. 2019: 603-612.

[24] Katsamenis I, Protopapadakis E, Doulamis A, et al. Pixel-level corrosion detection on metal constructions by fusion of deep learning semantic and contour segmentation[C]// Advances in Visual Computing: 15th International Symposium, ISVC 2020, San Diego, CA, USA, October 5–7, 2020, Proceedings, Part I. Cham: Springer International Publishing, 2020: 160-169.

[25] Liu J, Yang X, Lau S, et al. Automated pavement crack detection and segmentation based on two-step convolutional neural network[J]. Computer-Aided Civil and Infrastructure Engineering, 2020, 35(11): 1291-1305.

[26] Pan G, Zheng Y, Guo S, et al. Automatic sewer pipe defect semantic segmentation based on improved U-Net[J]. Automation in Construction, 2020, 119(29): 103383.

[27] 陈法法, 成孟腾, 杨蕴鹏, 等. 融合双注意力机制和U-Net网络的锈蚀图像分割[J]. 西安交通大学学报, 2021, 55(12): 119-128.

[28] Ye Z, Wei J, Lin Y, et al. Extraction of olive crown based on UAV Visible images and the U2-Net deep learning model[J]. Remote Sensing, 2022, 14(6): 1523.

[29] Gu J, Wang Z, Kuen J, et al. Recent advances in convolutional neural networks[J]. Pattern recognition, 2018, 77(51): 354-377.

[30] Ghosh A, Sufian A, Sultana F, et al. Fundamental concepts of convolutional neural network[J]. Recent trends and advances in artificial intelligence and Internet of Things, 2020: 519-567.

[31] Sharma S, Sharma S, Athaiya A. Activation functions in neural networks[J]. Towards Data Sci, 2017, 6(12): 310-316.

[32] He K, Zhang X, Ren S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J]. IEEE transactions on pattern analysis and machine intelligence, 2015, 37(9): 1904-1916.

[33] Zhou Z, He Z, Jia Y. AFPNet: A 3D fully convolutional neural network with atrous-convolution feature pyramid for brain tumor segmentation via MRI images[J]. Neurocomputing, 2020, 402(32): 235-244.

[34] Sifre L, Mallat S. Rigid-motion scattering for texture classification[J]. arXiv preprint arXiv:1403.1687, 2014.

[35] Liu S, Huang D. Receptive field block net for accurate and fast object detection[C]// Proceedings of the European conference on computer vision (ECCV). 2018: 385-400.

[36] 任欢, 王旭光. 注意力机制综述[J]. 计算机应用, 2021, 41(S1): 1-6.

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

[38] Jaderberg M, Simonyan K, Zisserman A. Spatial transformer networks[J]. Advances in neural information processing systems, 2015, 28: 2017–2025.

[39] Woo S, Park J, Lee J Y, et al. Cbam: Convolutional block attention module[C]// Proceedings of the European conference on computer vision (ECCV). 2018, 11211: 3-19.

[40] Qin X, Zhang Z, Huang C, et al. U2-Net: Going deeper with nested U-structure for salient object detection[J]. Pattern recognition, 2020, 106(53): 107404.

[41] 李翠云, 白静, 郑凉. 融合边缘增强注意力机制和U-Net网络的医学图像分割[J]. 图学学报, 2022, 43(2): 273-278.

[42] 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, Germany, October 5-9, 2015, Proceedings, Part III 18. Springer International Publishing, 2015: 234-241.

[43] Lin T Y，Goyal P，Girshick R，et al．Focal loss for dense object detection[C]// Proceedings of the IEEE International Conference on Computer Vision. Venice: IEEE，2017: 2980-2988．

[44] Salehi S S M，Erdogmus D，Gholipour A. Tversky loss function for image segmentation using 3D fully convolutional deep networks[C]// International Workshop on Machine Learning in Medical Imaging．Cham: Springer，2017: 379-387．

[45] Yu F, Koltun V. Multi-scale context aggregation by dilated convolutions[J]. arXiv preprint arXiv:1511.07122, 2015.

[46] Szegedy C，Vanhoucke V，Ioffe S，et al． Rethinking the inception architecture for computer vision[C]// Proceedings of the IEEE conference on Computer Vision and Pattern Recognition． New York，USA: IEEE，2016: 2818-2826．

[47] Yin B, Josselyn N, Considine T, et al. Corrosion Image Data Set for Automating Scientific Assessment of Materials [EB/OL]. British Machine Vision Conference (BMVC). (2021-09-20)[2023-05-15]. https://arl.wpi.edu/corrosion_dataset/#

[48] Badrinarayanan V, Kendall A, Cipolla R. Segnet: A deep convolutional encoder-decoder architecture for image segmentation[J]. IEEE transactions on pattern analysis and machine intelligence, 2017, 39(12): 2481-2495.

[49] Haghighi S, Jasemi M, Hessabi S, et al. PyCM: Multiclass confusion matrix library in Python[J]. Journal of Open Source Software, 2018, 3(25): 729.

[50] Geneva, CH. Paints and varnishes-Evaluation of degradation of coatings-Designation of quantity and size of defects, and of intensity of uniform changes in appearance-Part 8: Assessment of degree of delamination and corrosion around a scribe or other artifificial defect, International Organization for Standardization. Standard, Nov. 2012.

[51] 王文帝. 自旋转式图像法纱线毛羽数字化检测及其应用[D]. 上海: 上海工程技术大学, 2020.

[52] Huang J, Yang R, Ge H, et al. An effective determination of the minimum circumscribed circle and maximum inscribed circle using the subzone division approach[J]. Measurement Science and Technology, 2021, 32(7): 075014.

[53] 刘娟红, 孟翔, 段品佳, 等. 基于MATLAB的混凝土裂缝宽度计算方法研究[J]. 材料导报, 2022, 36(06): 84