基于深度学习的西北地区临水混凝土病害智能识别

Deep-Learning-Based Intelligent Recognition of Defects in Waterside Concrete Structures in Northwest China

摘要

针对西北地区临水钢筋混凝土结构病害检测存在的高成本、低效率及主观性强等问题，提出了一种基于卷积神经网络（CNN）的多阶段智能识别方法。通过融合像素级调整、空间域变换与几何变换的混合数据增强策略，将 550 张原始图像扩增至 3 840 张，有效缓解了实际工程中小样本数据的泛化问题。以 EfficientNet-B3 为主干网络，引入 EnhancedDyConvSE 模块中的增强型动态卷积与 SE 注意力机制，并结合混合权重方法，显著提升了对多样化病害形态特征的提取能力。结果表明，该模型在自建临水结构病害数据集上取得了 98.18% 的准确率与 98.20% 的精确率，能够准确识别盐结晶、坑蚀、剥落、露筋锈蚀和其他病害等 5 类典型病害，其性能显著优于 ResNet-50、DenseNet-121 等主流网络。

Abstract

To address the high cost, low efficiency, and strong subjectivity associated with defect detection in waterside reinforced-concrete structures in Northwest China, a multistage intelligent recognition method based on a convolutional neural network (CNN) was proposed. A hybrid data augmentation strategy integrating pixel-level adjustments, spatial-domain transformations, and geometric transformations was employed to expand the original dataset from 550 to 3,840 images, effectively alleviating the poor generalization associated with small datasets in practical engineering applications. With EfficientNet-B3 as the backbone network, the EnhancedDyConvSE module, incorporating enhanced dynamic convolution and a squeeze-and-excitation (SE) attention mechanism, was introduced and combined with a hybrid weighting method. This approach significantly enhanced the model’s ability to extract morphological features of diverse defects. The results showed that the model achieved an accuracy of 98.18% and a precision of 98.20% on a self-constructed dataset of defects in waterside structures. It accurately recognized five typical types of defects, including salt crystallization, pitting, spalling, and exposed and corroded reinforcement, and significantly outperformed mainstream networks such as ResNet-50 and DenseNet-121.

关键词

混凝土病害 / 图像扩增 / 卷积神经网络（CNN） / 图像识别 / 智能检测

Key words

concrete defects / image augmentation / convolutional neural network (CNN) / image recognition / intelligent detection

参考文献

[1] 宋俊杰. 宁夏地区在役临水钢筋混凝土柱腐蚀现状调研和剩余极限承载力预测研究[D]. 银川： 宁夏大学， 2022.

[2] 杨扬， 王连发， 张宇峰， 等. 基于多特征融合的混凝土结构表面病害图像分类算法[J]. 长安大学学报（自然科学版）， 2021， 41（3）： 64-74.

[3] 唐钱龙， 谭园， 彭立敏， 等. 基于数字图像技术的隧道衬砌裂缝识别方法研究[J]. 铁道科学与工程学报， 2019， 16（12）： 3041-3049.

[4] 谢明志， 樊丁萌， 蒋志鹏， 等.基于计算机视觉的混凝土结构裂缝检测研究现状与展望[J/OL]. 西南交通大学学报， 1-20[2026-07-24]. https://link.cnki.net/urlid/51.1277.u.20240913.1853.002.

[5] 刘鸿达， 孙旭辉， 李沂滨， 等. 基于卷积神经网络的图像分类深度学习模型综述[J]. 计算机工程与应用， 2025， 61（11）： 1-21.

[6] Chen L K， Chen W X， Wang L， et al. Convolutional neural networks （CNNs）-based multi-category damage detection and recognition of high-speed rail （HSR） reinforced concrete （RC） bridges using test images[J]. Engineering Structures， 2023， 276： 115306.

[7] Han K， Wang Y H， Chen H T， et al. A survey on vision transformer[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2023， 45（1）： 87-110.

[8] 蔡体健， 刘伟 . 数字图像处理： 基于 Python[M]. 北京： 机械工业出版社， 2022： 54-130.

[9] He K M， Zhang X Y， Ren S Q， et al. Deep residual learning for image recognition[C]//IEEE Conference on Computer Vision and Pattern Recognition （CVPR）. Piscataway，NJ： IEEE， 2016： 770-778.

[10] Huang G， Liu Z， Van Der Maaten L， et al. Densely connected convolutional networks[C]//IEEE Conference on Computer Vision and Pattern Recognition （CVPR）. Piscataway，NJ： IEEE， 2017： 2261-2269.

[11] Howard A G， Zhu M L， Chen B， et al. MobileNets： Efficient convolutional neural networks for mobile vision applications[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition （CVPR）. Piscataway，NJ： IEEE， 2017： 1-9.

[12] Tan M X， Le Q V. EfficientNet： Rethinking model scaling for convolutional neural networks[C]//Proceedings of the 36th International Conference on Machine Learning. Long Beach. PMLR， 2019： 6105-6114.

[13] Feng D C， Yi X， Deger Z T， et al. Rapid post-earthquake damage assessment of building portfolios through deep learning-based component-level image recognition [J]. Journal of Building Engineering， 2024， 98： 111380.

[14] Chen Y P， Dai X Y， Liu M C， et al. Dynamic convolution： Attention over convolution kernels [C]// IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR）. Piscataway，NJ： IEEE， 2020： 11027-11036.

[15] Hu J， Shen L， Sun G. Squeeze-and-excitation networks[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway，NJ： IEEE， 2018： 7132-7141.

[16] Malarvannan S， V P， Venkatraman S， et al. A novel adaptive hybrid focal-entropy loss for enhancing diabetic retinopathy detection using convolutional neural networks[PP/OL]. V2. arXiv （2025-04-23）[2026-07-24]. https://doi.org/10.48550/arXiv.2411.10843.

[17] Lin T Y， Goyal P， Girshick R， et al. Focal loss for dense object detection[C]// IEEE International Conference on Computer Vision （ICCV）. Piscataway， NJ： IEEE， 2017： 2999-3007.

[18] Aminudin M A I， Abdullah M N， Mustapha F， et al. Explainable deep learning framework for binary corrosion image classification using grad-CAM[J]. Sensors， 2025， 25（22）： 7070.

[19] Wang Q L， Wu B G， Zhu P F， et al. ECA-net： Efficient channel attention for deep convolutional neural networks[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition （CVPR）. Piscataway， NJ： IEEE， 2020： 11531-11539.

[20] Yang L X， Zhang R Y， Li L D， et al. SimAM： A simple， parameter-free attention module for convolutional neural networks[C]//International Conference on Machine Learning. Los Angeles. PMLR， 2021： 11863-11874.

基金资助

宁夏自然科学基金重点项目(2025AAC020028)