基于轻量级卷积神经网络MobileNetV2的混凝土结构裂缝识别

范立军

doi:10.13204/j.gyjzG23021618

基于轻量级卷积神经网络MobileNetV2的混凝土结构裂缝识别

doi: 10.13204/j.gyjzG23021618

范立军^1,2

1. 中铁建设集团有限公司, 北京 100043;
2. 中铁建设集团有限公司工程研究院, 北京 100043

基金项目:

海南省科技厅重点研发项目（ZDYF2022SHFZ337）。

详细信息

作者简介:
范立军,男,1975年出生,高级工程师。电子信箱:459013833@qq.com

计量
- 文章访问数: 96
- HTML全文浏览量: 12
- PDF下载量: 4
- 被引次数: 0
出版历程
- 收稿日期: 2023-02-16

Identification of Crack in Concrete Structures Based on MobileNetV2 of Lightweight Convolutional Network

FAN Lijun^1,2

1. China Railway Construction Group Co., Ltd., Beijing 100043, China;
2. Engineering Research Institute, China Railway Construction Group Co., Ltd., Beijing 100043, China

摘要

摘要: 混凝土结构随着服役时间的增长，产生的裂缝会不断扩展并可能对结构造成损伤。因此，裂缝检测对于混凝土结构健康监测有着重要意义，但超声及基于人工视觉等传统检测方法无法大量快速地对裂缝进行分类检测。为此提出一种基于移动网络（MobileNetV2）轻量级卷积网络和谷歌张量流图（TensorFlow）深度学习框架的混凝土结构裂缝快速识别分类的预测模型。首先，基于路径库存（pathlib）方法对数据进行提取并划分数据集；其次，基于迁移学习的数据增强对数据集进行扩充；再次，在TensorFlow框架下利用深度学习的接口（Keras）来构建卷积网络的池化层等；最后，建立构建完整卷积网络架构并输出结果。试验结果表明：模型收敛时，预测精度达到0.997 5，训练时间仅为710 s，可以为工程现场的移动设备检测提供帮助。
- 深度学习 /
- 卷积神经网络 /
- 混凝土结构 /
- MobileNetV2 /
- 裂缝
Abstract: With the increase of service time of concrete structures, the generated cracks will continue to expand and may cause damage to the structure. Therefore, crack detection is of great significance for the health monitoring of concrete structures, but traditional ultrasonic and artificial vision-based detection methods can not quickly classify cracks. Based on MobileNetV2 lightweight convolutional network and TensorFlow deep learning framework, a prediction model for rapid identification and classification of concrete structure crack was established. Firstly, the data set was extracted and partitioned based on the pathlib method; secondly, data enhancement based on transfer learning expanded the data set; thirdly, based on the TensorFlow framework, Keras was used to build the pooling layer of the convolutional network; finally, a complete convolutional network architecture and obtained the results were obtained. The test results showed that the model converged, the prediction accuracy reached 0.997 5, and the training time was only 710 s. It could provide help for mobile equipment detection in the project site.
- deep learning /
- convolutional network /
- concrete structures /
- MobileNetV2 /
- crack

HTML全文

参考文献(18)

[1]	ZHANG L, YANG F, ZHANG Y D, et al. Road crack detection using deep convolutional neural network[C]//IEEE International Conference on Image Processing (ICIP). 2016:3708-3712.
[2]	DUNG C V, ANH L D. Autonomous concrete crack detection using deep fully convolutional neural network[J]. Automation in Construction, 2019, 99:52-58.
[3]	鲍跃全,李惠. 人工智能时代的土木工程[J]. 土木工程学报, 2019, 52(5):1-11.
[4]	ZHAO J, LI L. Research on image classification algorithm based on convolutional neural Network-TensorFlow[J/OL]. Journal of Physics,2010[2022-08-11].https://iopscience.iop.org/article/10.1088/1742-6596/2083/3/032054.
[5]	GU J, WANG Z, KUEN J, et al. Recent advances in convolutional neural networks[J]. Pattern Recognition, 2018,77:354-377, 3.
[6]	HE T, LIU Y, YU Y, et al. Application of deep convolutional neural network on feature extraction and detection of wood defects Çaǧlar Flrat Özgenel:Concrete Crack Images for Classification[J/OL]. Measurement, 2020[2023-02-11]. https://doi.org/10.1016/j.measurement.2019.107357.
[7]	LI W, FIELD K G, MORGAN D. Automated defect analysis in electron microscopic images[J]. NPJ Computational Materials, 2018(4):1-9.
[8]	DONG Y, SU C, QIAO P, et al. Microstructural crack segmentation of three-dimensional concrete images based on deep convolutional neural networks[J/OL]. Construction and Building Materials, 2020[2022-08-11].https://doi.org/10.1016/j.conbuildmat.2020.119185.
[9]	DAS A K, LEUNG C K Y, WAN K T. Application of deep convolutional neural networks for automated and rapid identification and computation of crack statistics of thin cracks in strain hardening cementitious composites (SHCCs)[J/OL]. Cement and Concrete Composites, 2021[2022-08-11].https://doi.org/10.1016/j.cemconcomp.2021.104159.
[10]	REZAIE A, ACHANTA R, GODIO M, et al. Comparison of crack segmentation using digital image correlation measurements and deep learning[J/OL]. Construction and Building Materials, 2020[2022-08-11].https://doi.org/10.1016/j.conbuildmat.2020.120474.
[11]	OH B K, PARK H S, GLISIC B. Prediction of long-term strain in concrete structure using convolutional neural networks, air temperature and time stamp of measurements[J/OL]. Automation in Construction, 2021[2022-08-11].https://doi.org/10.1016/j.autcon.
[12]	KHODABANDEHLOU H, PEKCAN G, FADALI M S. Vibration-based structural condition assessment using convolution neural networks[J/OL]. Structural Control and Health Monitoring,2018[2018-12-13].https://doi.org/10.1002/stc.2308.
[13]	WANG L, KAWAGUCHI K, WANG P. Damaged ceiling detection and localization in large-span structures using convolutional neural networks[J/OL]. Automation in Construction,2020[2022-08-11].https://doi.org/10.1016/j.autcon.2020.103230.
[14]	KUMAR N, RATHEE M, CHANDRAN N, et al. CrypTFlow:Secure TensorFlow inference[C]//Proceedings of the IEEE Symposium on Security and Privacy. 2020:336-353.
[15]	HOWARD A G, ZHU M, CHEN B, et al. MobileNets:efficient convolutional neural networks for mobile vision applications[EB/OL].[2022-08-11]. http://arxiv.org/abs/1704.04861.
[16]	ÖZGENEL F, GÖNENÇ S A. Performance comparison of pretrained convolutional neural networks on crack detection in buildings[C]//Proceedings of the ISARC 2018:35th International Symposium on Automation and Robotics in Construction and International AEC/FM Hackathon:The Future of Building Things. Berlin:2018.
[17]	SANDLER M, HOWARD A, ZHU M, et al. MobileNet V2:Inverted residuals and linear bottlenecks[C]//Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition.New York:IEEE, 2018:4510-4520.
[18]	Çaǧlar Flrat Özgenel:Concrete crack images for classification[R/OL].Berlin:Orta Dogu Teknik Universitesi, 2019[2019-07-23]. https://data.mendeley.com/datasets/5y9wdsg2zt/2.