One-Class Detection and Classification of Defects on Concrete Surfaces

Today, railway infrastructure is subject to regular inspections carried out by image acquisition. Experts need automatic tools to process these images and extract defects. These defects can be classified in two categories : linear defects, like cracks, and surface defects, like spall, humidity, graf...

Full description

Saved in:
Bibliographic Details
Published inConference proceedings - IEEE International Conference on Systems, Man, and Cybernetics pp. 826 - 831
Main Authors Faula, Yannick, Eglin, Veronique, Bres, Stephane
Format Conference Proceeding
LanguageEnglish
Published IEEE 17.10.2021
Subjects
Detectors
Generative adversarial networks
Humidity
Inspection
Neural networks
Rail transportation
Support vector machines
Online AccessGet full text

Cover

More Information
Summary:Today, railway infrastructure is subject to regular inspections carried out by image acquisition. Experts need automatic tools to process these images and extract defects. These defects can be classified in two categories : linear defects, like cracks, and surface defects, like spall, humidity, graffiti... Cracks detection has been presented in a previous work, and is based on Local Binary Patterns (LBP) extraction and our FLASH algorithm [1]. In this paper, we propose a new method, using the same LBP descriptors, already extracted, to detect surface defects using generative adversarial networks (GAN). The detection is followed by a classification with a potentially scalable system. Experiments show that our detector achieves better performance than the state-of-the art in application to defect detection.
ISSN:2577-1655
DOI:10.1109/SMC52423.2021.9659221