| 399 | 0 | 107 |
| 下载次数 | 被引频次 | 阅读次数 |
针对常规桥梁定期检查中移动性不足、硬件资源受限及过度依赖现场作业等问题，提出一种基于YOLOv5的轻量化桥梁表观病害检测方法。在YOLOv5算法框架基础上，分别结合轻量型网络架构ShuffleNetV2和MobileNetV3的优点进行轻量化改进。首先，在保留YOLOv5特征融合模块的基础上，引入ShuffleNetV2网络并在网络末端嵌入SE注意力机制，以补偿模型简化带来的精度损失。其次，用MobileNetV3网络的特征提取核心结构Bneck替换YOLOv5的主干网络，得到YOLOv5-MNv3模型，减少了模型参数和计算量。最后，基于常规桥梁表观病害数据集进行试验，试验结果表明：改进后ShuffleNetV2性能大幅提升，参数量和计算量分别减少36.10%和45.57%,但模型平均精度(mAP)与检测准确率下降显著。MobileNetV3核心结构的轻量化模型保持原有mAP值的前提下，检测精确度仅降1.01%,参数量和计算量分别降低46.8%和58.9%。
Abstract：A lightweight bridge apparent disease detection method based on YOLOv5 is proposed to address the problems of mobility requirements, limited hardware resources and over-reliance on field operations in regular bridge periodic inspections. Based on the YOLOv5 algorithm framework, the lightweight improvement is made by combining the advantages of lightweight network architectures ShuffleNetV2 and MobileNetV3, respectively. Firstly, on the basis of retaining the YOLOv5 feature fusion module, the ShuffleNetV2 network is introduced and the SE attention mechanism is embedded at the end of the network to compensate for the loss of accuracy due to model simplification. Secondly, the feature extraction core structure Bneck of MobileNetV3 network is used to replace the backbone network of YOLOv5 to obtain the YOLOv5-MNv3 model, which reduces the model parameters and computation. Finally, experiments are conducted based on the conventional bridge apparent disease dataset, and the experimental results show that: the performance of ShuffleNetV2 is greatly improved after the improvement, and the number of parameters and computation volume are reduced by 36.10% and 45.57%, respectively, but the model mean average precision(mAP) and the detection accuracy drop significantly. The lightweight model of MobileNetV3 core structure maintains the original mAP value, the detection accuracy is only reduced by 1.01%, and the number of parameters and computation amount are reduced by 46.8% and 58.9%, respectively.
[1] 樊健生，刘宇飞.在役桥梁检测、健康监测技术现状与时空融合诊断体系研究[J].市政技术，2022,40(8):1-11,40.
[2] 高俊祥.基于图像处理和机器学习的桥梁检测新技术研究[D].南京：东南大学，2018.
[3] 唐永，么学春，王晨，等.基于无人机图像技术和支持向量机(SVM)的桥梁裂缝自动识别系统[J].公路工程，2024,49(6):49-56.
[4] 刘宇飞，冯楚乔，陈伟乐，等.基于机器视觉法的桥梁表观病害检测研究综述[J].中国公路学报，2024,37(2):1-15.
[5] JIANG S,CHENG Y Y,ZHANG J.Vision-guided unmanned aerial system for rapid multiple-type damage detection and localization[J].Structural Health Monitoring,2023,22(1):319-337.
[6] KIM B,NATARAJAN Y,PREETHAA K R S,et al.Real-time assessment of surface cracks in concrete structures using integrated deep neural networks with autonomous unmanned aerial vehicle[J].Engineering Applications of Artificial Intelligence,2024,129:107537.
[7] AMIEGHEMEN G E,SHERIF M M.Deep convolutional neural network ensemble for pavement crack detection using high elevation UAV images[J].Structure and Infrastructure Engineering,2025,21(6):1008-1023.
[8] 魏法胜，高亮，钟运平，等.基于无人机的曲线宽幅矮塔斜拉桥表观病害智能检测[J].公路，2023,68(10):155-160.
[9] 阮小丽，钟建平，吴巨峰，等.基于无人机的桥梁外露面裂缝识别系统研究[J].湖南交通科技，2023,49(3):104-108,114.
[10] YANG L,LI B,FENG J L,et al.Automated wall-climbing robot for concrete construction inspection[J].Journal of Field Robotics,2023,40(1):110-129.
[11] 杨正.基于机器人的桥梁检测病害识别和路径规划研究[D].长春：吉林大学，2022.
[12] LYU G Z,WANG P,LI G H,et al.A heavy-load wall-climbing robot for bridge concrete structures inspection[J].Industrial Robot,2024,51(3):465-478.
[13] YEUM C M,DYKE S J.Vision-based automated crack detection for bridge inspection[J].Computer-Aided Civil and Infrastructure Engineering,2015,30(10):759-770.
[14] SUN L M,SHANG Z Q,XIA Y,et al.Review of bridge structural health monitoring aided by big data and artificial intelligence:from condition assessment to damage detection[J].Journal of Structural Engineering,2020,146(5):04020073.
[15] ZHANG C B,CHANG C C,JAMSHIDI M.Concrete bridge surface damage detection using a single-stage detector[J].Computer-Aided Civil and Infrastructure Engineering,2020,35(4):389-409.
[16] 王超，贾贺，张社荣，等.基于图像的混凝土表面裂缝量化高效识别方法[J].水力发电学报，2021,40(3):134-144.
[17] WU P R,LIU A R,FU J Y,et al.Autonomous surface crack identification of concrete structures based on an improved one-stage object detection algorithm[J].Engineering Structures,2022,272:114962.
[18] 刘德坤，刘肖亮，张帅.基于YOLOv5的路面病害图像识别方法研究[J].公路工程，2024,49(3):66-75.
[19] 董绍江，谭浩，刘超，等.改进YOLOv5s的桥梁表观病害检测方法[J].重庆大学学报，2024,47(9):91-100.
[20] LI R X,YU J Y,LI F,et al.Automatic bridge crack detection using unmanned aerial vehicle and Faster R-CNN[J].Construction and Building Materials,2023,362:129659.
[21] 张振海，季坤，党建武.基于桥梁裂缝识别模型的桥梁裂缝病害识别方法[J].吉林大学学报(工学版),2023,53(5):1418-1426.
[22] ZHANG X Y,ZHOU X Y,LIN M X,et al.ShuffleNet:an extremely efficient convolutional neural network for mobile devices[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE,2017:6848-6856.
[23] MA N N,ZHANG X Y,ZHENG H T,et al.ShuffleNet V2:practical guidelines for efficient CNN architecture design[C]//FERRARI V,HEBERT M,SMINCHISESCU C,et al.Computer Vision—ECCV 2018.Cham:Springer,2018:122-138.
[24] HU J,SHEN L,SUN G.Squeeze-and-excitation networks[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE,2018:7132-7141.
[25] HOWARD A G,ZHU M L,CHEN B,et al.MobileNets:efficient convolutional neural networks for mobile vision applications[DB/OL].(2017-04-17)[2024-10-15].https://arxiv.org/abs/1704.04861.
[26] SANDLER M,HOWARD A,ZHU M L,et al.MobileNetV2:inverted residuals and linear bottlenecks[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.Piscataway:IEEE,2018:4510-4520.
[27] MUNDT M,MAJUMDER S,MURALI S,et al.Meta-Learning convolutional neural architectures for multi-target concrete defect classification with the concrete defect bridge image dataset[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR).Piscataway:IEEE,2019:11188-11197.
基本信息:
DOI：10.19782/j.cnki.1674-0610.2025.06.003
中图分类号:U445.71
引用信息:
[1]黄蓉,李睿,翟雨辰,等.基于YOLOv5的桥梁表观病害轻量化检测方法[J].公路工程,2025,50(06):23-31+103.DOI:10.19782/j.cnki.1674-0610.2025.06.003.
基金信息:
国家自然科学基金项目（52068037）
2024-11-28
2024
2025-04-28
2025
2
2025-05-22
2025-05-22
2025-05-22