基于递进式级联卷积神经网络的混凝土裂缝识别方法

卢佳祁; 姚志东

doi:10.13204/j.gyjzG20112504

基于递进式级联卷积神经网络的混凝土裂缝识别方法

doi: 10.13204/j.gyjzG20112504

卢佳祁^1,2,
姚志东^1,2

1. 中冶建筑研究总院(深圳)有限公司, 广东深圳 518055;
2. 深圳市建筑幕墙智能检测工程技术研究中心, 广东深圳 518055

基金项目:

国家重点研发计划项目（2019YFB1600700）；中冶建筑研究总院有限公司重大课题项目（YJA2018Kj01）。

详细信息

作者简介:
卢佳祁,男,1988年出生,硕士,工程师。电子信箱:15104573199@163.com

计量
- 文章访问数: 240
- HTML全文浏览量: 15
- PDF下载量: 8
- 被引次数: 0
出版历程
- 收稿日期: 2020-11-25
- 网络出版日期: 2021-09-16
- 刊出日期: 2021-09-16

A CONCRETE CRACK RECOGNITION METHOD BASED ON PROGRESSIVE CASCADE CONVOLUTION NEURAL NETWORK

LU Jiaqi^1,2,
YAO Zhidong^1,2

1. Central Research Institute of Building and Construction (Shenzhen) Co., Ltd., Shenzhen 518055, China;
2. Shenzhen Engineering Research Center of Intelligent Inspection for Curtain Wall, Shenzhen 518055, China

摘要

摘要: 基于深度学习的卷积神经网络方法是目前图像裂缝识别鲁棒性较高的方法，主要分为滑动窗口法和图像分割法。滑动窗口法存在后期阈值分割裂缝精度不高的问题；全局图像分割法存在裂缝区域数据和背景区域数据严重不均衡问题，会对裂缝分割精度产生影响。采用了基于递进式级联卷积神经网络的方法对混凝土表面裂缝进行识别：首先采用全卷积神经网络一次性判断图像中所有密集重叠窗口区域内是否含有裂缝，然后将含有裂缝的窗口区块提取出来作为感兴趣区域，再基于轻量化的U-Net图像分割网络作用于感兴趣区域，将裂缝区域精确地提取出来。试验结果表明，所提出的基于递进式级联卷积神经网路的裂缝识别方法优于直接使用滑动窗口法和全局图像分割法，有着可靠的应用前景。
- 裂缝识别 /
- 递进式级联卷积神经网络 /
- 全卷积神经网络 /
- 感兴趣区域 /
- U-Net图像分割
Abstract: Convolution neural network method of deep learning is a high robust method for image crack recognition at present, which is mainly divided into sliding window method and image segmentation method. Sliding window method has the problems of low precision of later threshold segmentation of cracks; global image segmentation method has the problem of serious unbalanced sample distribution between crack region and background region,which will affect the accuracy of crack segmentation. The method based on progressive cascade convolution neural network was used to detect concrete surface cracks:firstly, the fully convolution neural network was used to judge whether there were cracks in all the dense overlapped window areas in the image only once, and then the window blocks with cracks were extracted as the region of interest, and then the light-weight U-Net image segmentation network was used to act on the region of interest to extract the crack area accurately. Experimental results showed that the proposed progressive cascade convolution neural network was superior to sliding window method and global image segmentation method, and had a reliable application prospect.
- crack recognition /
- progressive cascade convolution neural network /
- fully convolutional network /
- region of interest /
- U-Net image segmentation

HTML全文

参考文献(20)

[1]	FENG D, FENG M Q. Computer Vision for SHM of Civil Infrastructure:from Dynamic Response Measurement to Damage Detection:A Review[J]. Engineering Structures, 2018, 156:105-117.
[2]	KONG X, LI J. Non-Contact Fatigue Crack Detection in Civil Infrastructure Throughimage Overlapping and Crack Breathing Sensing[J]. Automation in Construction, 2019, 99:125-139.
[3]	OLIVEIRA H, CORREIA P L. Automatic Road Crack Segmentation Using Entropy and Image Dynamic Thresholding[C]//Proceedings of the 2009 17th European Signal Processing Conference. Glasgow, U K:2009:622-626.
[4]	QUINTANA M, TORRES J, MENéNDEZ J M. A Simplified Computer Vision System for Road Surface Inspection and Maintenance[J]. IEEE Transactions on Intelligent Transportation Systems, 2016, 17(3):608-619.
[5]	AYENU-PRAH A, ATTOH-OKINE N.Evaluating Pavement Cracks with Bidimensional Empirical Mode Decomposition[J]. EURASIP Journal on Advances in Signal Processing, 2008(1). DOI: 10.1155/2008/861701.
[6]	ACHARYA T, TSAI P S. Edge-Detection Based Noise Removal Algorithm:US 6229578 B1[P]. 1997-12-08.
[7]	LI Q Q, LIU X L.Novel Approach to Pavement Image Segmentation Based on Neighboring Difference Histogram Method[C]//Proceedings of the 2008 Congress on Image and Signal Processing.Sanya:2008:792-796.
[8]	JIN H Z, WAN F, YE Z W. Pavement Crack Detection Fused HOG and Watershed Algorithm of Range image[J]. Journal of Huazhong Normal University(Natural Sciences), 2017, 51(5):715-722.
[9]	徐洋. 基于计算机视觉的桥梁结构局部损伤识别方法研究[D]. 哈尔滨:哈尔滨工业大学, 2019.
[10]	李良福, 马卫飞, 李丽, 等. 基于深度学习的桥梁裂缝检测算法研究[J]. 自动化学报, 2019, 45(9):1727-1742.
[11]	王森, 伍星, 张印辉, 等. 基于深度学习的全卷积网络图像裂纹检测[J]. 计算机辅助设计与图形学学报, 2018, 30(5):859-867.
[12]	曹锦纲, 杨国田, 杨锡运, 等. 基于注意力机制的深度学习路面裂缝检测[J]. 计算机辅助设计与图形学报, 2020, 32(8):1324-1333.
[13]	任秋兵, 李明超, 沈扬, 等.水工混凝土裂缝像素级形态分割与特征量化方法研究[J]. 水力发电学报, 2021, 40(2):234-246.
[14]	OTSU N. A Threshold Selection Method from Gray-Level Histogram[J]. IEEE Transactions on Systems, Man, and Cybernetics, 1979, 9(1):62-66.
[15]	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.
[16]	GONZALEZ R C, WOODS R E, EDDINS S L. 数字图像处理的MATLAB实现[M]. 阮秋琦, 译.北京:清华大学出版社, 2013.
[17]	李刚, 贺拴海, 巨永锋, 等. 远距离混凝土桥梁结构表面裂缝精确提取算法[J]. 中国公路学报, 2013, 26(4):102-108.
[18]	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. 2017:2980-2988.
[19]	RONNEBERGER O, FISCHER P, BROX T. U-Net:Convolutional Networks for Biomedical Image Segmentation[C]//International Conference on Medical Image Computing and Computer-Assisted Intervention.2015:234-241.
[20]	SERMANET P, EIGEN D, ZHANG X, et al. OverFeat:Integrated Recognition, Localization and Detection Using Convolutional Networks[C]//International Conference on Learning Representations.2014:1-16.