Crack detection via salient structure extraction from textured background

Nayyeri, F, Hou, L, Zhou, J, Guan, H and Liew, A 2018, 'Crack detection via salient structure extraction from textured background', in Tommy Chan, Saeed Mahini (ed.) Proceedings of the 8th International Conference on Structural Health Monitoring of Intelligent Infrastructure (SHMII-8), Brisbane, Australia, 5-8 December 2017, pp. 1-8.


Document type: Conference Paper
Collection: Conference Papers

Title Crack detection via salient structure extraction from textured background
Author(s) Nayyeri, F
Hou, L
Zhou, J
Guan, H
Liew, A
Year 2018
Conference name SHMII-8: Structural Health Monitoring in Real-world Application
Conference location Brisbane, Australia
Conference dates 5-8 December 2017
Proceedings title Proceedings of the 8th International Conference on Structural Health Monitoring of Intelligent Infrastructure (SHMII-8)
Editor(s) Tommy Chan, Saeed Mahini
Publisher Queensland University of Technology
Place of publication Brisbane, Australia
Start page 1
End page 8
Total pages 8
Abstract A reliable crack detection system is essential for automatic safety inspection of infrastructures such as roads and bridges. Queenslands population is increasing, putting greater pressure on our already aging civil infrastructure. To get the most out of government investment in our States thousands of bridges, we need to find quicker, cheaper and more reliable ways to assess and maintain them. In this paper, we propose a novel method for crack detection via salient structure extraction from textured background. This method contains two key steps. In the first step, we extract strong edges and distinguish them from strong textures in a local neighborhood via a relative total variation approach. In the second step, the spatial distribution of texture features are calculated so as to detect cracks as salient structures that are not widely spread across the whole image. The outputs from these two steps are fused to calculate the final structure saliency map which is then binarised to generate the crack masks. This method was evaluated on a crack dataset with images collected from the Internet. Comparison with several alternative approaches shows the superior performance of our method.
Subjects Construction Engineering
Keyword(s) crack detection
civil infrastructure
salient structure extraction
texture feature
saliency map
Copyright notice ©
ISBN 9781925553055
Versions
Version Filter Type
Access Statistics: 65 Abstract Views  -  Detailed Statistics
Created: Wed, 19 Sep 2018, 13:35:00 EST by Catalyst Administrator
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us