Improved Global Convolutional Network for Pavement Crack Detection

Gang Li; Zhenyang Gao; Xinchun Zhang; Huaixin Zhao; Zhuo Liu

doi:10.3788/LOP57.081011

Laser & Optoelectronics Progress, Volume. 57, Issue 8, 081011(2020)

Improved Global Convolutional Network for Pavement Crack Detection

Gang Li^1、*, Zhenyang Gao¹, Xinchun Zhang¹, Huaixin Zhao², and Zhuo Liu³

Author Affiliations

¹School of Electronic and Control Engineering, Chang'an University, Xi'an, Shaanxi 710064, China

²Technology Quality Department, Shaanxi Province Railway Group Co., Ltd., Xi'an, Shaanxi 710199, China

³Commission for Discipline Inspection and Supervision, Xi'an Xilan Natural Gas Group Co., Ltd., Xi'an, Shaanxi 710075, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(18)

Fig. 1. Different network models. (a) Classification model; (b) segmentation model

Download full size

Fig. 2. Overall structure of the ResNet-GCN model. (a) Structure of the entire framework; (b) GCN structure; (c) boundary refinement module

Download full size

Fig. 3. Dataset containing different crack types. (a) Crack; (b) watery crack; (c) crack with repair seal; (d) crack with lane line; (e) stitching seam; (f) crack containing debris

Download full size

Fig. 4. Labeling of experimental crack data. (a) (b) (c) Original crack images; (d) (e) (f) manually marked cracks

Download full size

Fig. 5. Comparision of GCN and ordinary convolution kernel

Download full size

Fig. 6. Test accuracy of MobileNetv2-GCN model

Download full size

Fig. 7. Test m_IoU of MobileNetv2-GCN model

Download full size

Fig. 8. Crack segmentation effect of MobileNetv2-GCN model. (a) Original images; (b) label images; (c) prediction results

Download full size

Fig. 9. Crack skeleton extraction. (a)(b)(c) Binary images after segmentation; (d)(e)(f) extracted crack skeleton images

Download full size

Fig. 10. Comparison of real and predicted average crack width pixel

Download full size

Table 1. Crack dataset details
View table
Table 1. Crack dataset details
Data Training Verification Test
Image size /(pixel×pixel) 512× 512 512×512 512×512
Number of images 2160 720 720

Table 2. Comparison of GCN results with different convolution kernel sizes
View table
Table 2. Comparison of GCN results with different convolution kernel sizes
k Base 3 5 7 9 11 13 15
m_IoU /% 70.5 71.1 71.9 72.6 73.3 74.1 75.2 76.6

Table 3. Comparison of GCN and equivalent small kernel stack convolution
View table
Table 3. Comparison of GCN and equivalent small kernel stack convolution
k 3 5 7 9 11
m_IoU /% GCN 71.1 71.9 72.6 73.3 74.1
Stack 69.8 70.9 69.5 68.2 67.5

Table 4. Comparison of experimental results to reduce the number of stacked convolution layers
View table
Table 4. Comparison of experimental results to reduce the number of stacked convolution layers
Number oflayers Stack GCN
2048 1024 210 2048
m_IoU/% 71.1 70.5 68.6 72.8
Parameteramount /k 75885 28505 4307 608

Table 5. Comparison of experimental results after adding boundary refinement blocks
View table
Table 5. Comparison of experimental results after adding boundary refinement blocks
Model Boundaryaccuracy /% Centeraccuracy /% Overallaccuracy /%
Baseline 71.3 90.1 70.3
GCN 71.5 91.1 85.6
GCN+BR 72.6 92.3 86.7

Table 6. Comparison of experimental results of different pre-training fusion models
View table
Table 6. Comparison of experimental results of different pre-training fusion models
Model m_IoU /% Accuracy /% Model size /MB
ResNet- GCN 76.6 86.7 671.0
ResNet50-GCN 82.6 90.2 274.0
ResNet101-GCN 83.6 93.4 492.0
MobileNetv2-GCN 84.6 98.5 15.5

Table 7. Experimental results of MobileNetv2-GCN on different datasets
View table
Table 7. Experimental results of MobileNetv2-GCN on different datasets
Dataset Crack500 GAPs384 Cracktree200 Proposed
m_IoU /% 85.8 82.3 80.5 85.3
Accuracy /% 95.5 92.4 89.6 98.5

Table 8. Comparison of MobileNetv2-GCN and other crack segmentation models
View table
Table 8. Comparison of MobileNetv2-GCN and other crack segmentation models
Model SegNet DeepLab CNN FCN MobileNetv2-GCN
m_IoU /% 67.8 69.2 72.3 80.5 85.3
Accuracy /% 74.5 80.3 81.9 89.6 98.5

Tools

Get Citation

Copy Citation Text

Gang Li, Zhenyang Gao, Xinchun Zhang, Huaixin Zhao, Zhuo Liu. Improved Global Convolutional Network for Pavement Crack Detection[J]. Laser & Optoelectronics Progress, 2020, 57(8): 081011

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category: Image Processing

Received: Aug. 5, 2019

Accepted: Sep. 12, 2019

Published Online: Apr. 3, 2020

The Author Email: Li Gang (15229296166@chd.edu.cn)

DOI:10.3788/LOP57.081011

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology

Table 1. Crack dataset details

Table 1. Crack dataset details

Table 2. Comparison of GCN results with different convolution kernel sizes

Table 2. Comparison of GCN results with different convolution kernel sizes

Table 3. Comparison of GCN and equivalent small kernel stack convolution

Table 3. Comparison of GCN and equivalent small kernel stack convolution

Table 4. Comparison of experimental results to reduce the number of stacked convolution layers

Table 4. Comparison of experimental results to reduce the number of stacked convolution layers

Table 5. Comparison of experimental results after adding boundary refinement blocks

Table 5. Comparison of experimental results after adding boundary refinement blocks

Table 6. Comparison of experimental results of different pre-training fusion models

Table 6. Comparison of experimental results of different pre-training fusion models

Table 7. Experimental results of MobileNetv2-GCN on different datasets

Table 7. Experimental results of MobileNetv2-GCN on different datasets

Table 8. Comparison of MobileNetv2-GCN and other crack segmentation models

Table 8. Comparison of MobileNetv2-GCN and other crack segmentation models