Abstract:The use of weathering steel is increasingly important to prevent further corrosion of steel due to its special characteristics of forming a stable and protective oxide layer termed rust patina in suitable environments. As a result, the weathering steel having natural corrosion resistance is becoming an appropriate choice for structural components of new steel bridges since 1964. The monitoring and condition assessment for weathering steel bridges are becoming important aspects on appropriate material selection in planning and design stages, corrosion prediction and further maintenance planning during their service life. The popular NDT methods such as ion transfer resistance method, electrochemical potential method, X-ray diffraction method, spectroscopic identification techniques are employed to measure the corrosion behaviors from the point of view of corrosion loss, and chemic, electric and magnetic properties of the rust patina. Note that human vision has played an indispensable role in extracting effective and directive information from the external appearance. Since the atmospheric environment primarily causes the corrosion of the exterior of the weathering steel, the state of the surface rust is sufficient to exhibit the corrosion damage. Japan Association of Steel Bridge Construction and Japan Iron and Steel Federation proposed a standard criterion to evaluate the state of weathering steel based on visual inspection and a supplemental rust thickness measurement, and categorized the surface condition into 5 levels ranging from 5 to 1. It is obvious that the expert visual inspection and evaluation are subject to the impacts of subjective decision-making such as experience, ages, and even temper. Moreover, conventional visual analysis approaches are time-consuming work, having high degree of variability, failing to provide quantitative information. Recently, digital image recognition has been used experimentally in the automated system for bridge inspection, pavement management, sewer pipeline inspection, material engineering. The critical issues of this approach contain imaging-based feature extraction and pattern classification algorithm. For imaging-based feature extraction, wavelet transform (WT) has shown better performance in many cases due to its outstanding capabilities in space-frequency decomposition at each resolution. Another crucial issue is the pattern classification, numerous classifiers using the extracted features have been employed, including nearest neighbor classifier, Bayes classifier, probabilistic neural network, learning vector quantization. Recently, the support vector machine (SVM) is becoming a successful method for pattern recognition. In the cases of weathering steels, it is generally believed that the surface corrosion characteristics of them under different controlling factors can be described by morphology, i.e., color, texture and shape. If we are able to construct a relationship between morphological attributes of the training sample images and the corresponding qualified condition state, it is desirable to make a pattern classification on morphological attributes of the new test image. This paper presents an imaging-based method for condition rating of surface quality of the corroded weathering steel bridges using WT and SVM. A schematic of the imaging-based condition (Figure Presented) rating for weathering steel using wavelet texture analysis and SVM is shown in Figure 1. The surface image data of weathering steel acquired by digital camera were preprocessed to create standard rescaled grayscale image for wavelet texture analysis, from which both local and normalized global energy distribution of each detail sub-image are extracted as the feature vectors. Then, the SVMs were implemented to classify a set of wavelet texture features into various corrosion levels related to condition state of the surface images of weathering steel. The quality assessment was finally reported. The systematic investigation on a representative set of 530 images indicates that, by carefully select the decomposition level and texture feature vector, the averaged classification accuracy can attain to 82% in the case of using half samples for training and half for testing. The limitation of this imaging based pattern classification method is that the morphological characteristics of surface image such as color, texture, shape can cover most information of the surface quality but not all. Moreover, the critical feature of thickness of rust patina can not be fully indicated by morphology. The further study will placed emphases on selection of new feature vector and comparisons with many other pattern classification algorithms. © 2006 Taylor & Francis Group.

Imaging-based visual inspection for bridge condition assessment

Imaging-based surface quality assessment of weathering steel bridge based on wavelet transform and support vector machine

A new method for inspecting crack of concrete bridges using image processing technique

Automated damage detection of bridge’s sub-surface defects from infrared images using machine learning

Automatic Crack Inspection for Concrete Bridge Bottom Surfaces Based on Machine Vision

Computer Vision-Based Bridge Inspection and Monitoring: A Review

Imaging-Based Rating for Corrosion States of Weathering Steel Using Wavelet Transform and Pso-Svm Techniques

Steel Box-Girder Bridge Diseases Identification Based on Computer Vision System

Image-Based Comprehensive Maintenance and Inspection Method for Bridges Using Deep Learning

Review of Vibration-Based Damage Detection and Condition Assessment of Bridge Structures Using Structural Health Monitoring

Automated Change Diagnosis of Single Column Pier Bridges Based on 3D Imagery Data

Machine-Aided Bridge Deck Crack Condition State Assessment Using Artificial Intelligence

Application of Computer Vision Technology to Structural Health Monitoring of Engineering Structures

Study on the crack detection of bridges based on digital image processing

Computer Vision Techniques for Bridge Bearing Condition Assessment using Visual Inspection Photographs

An Automated Inspection Method for the Steel Box Girder Bottom of Long-Span Bridges Based on Deep Learning

An Approach for Auto Bridge Inspection Based on Climbing Robot

Measurement of Surface Crack Width Based on Digital Image Processing

Multi-Point Displacement Monitoring of Bridges Using A Vision-Based Approach

Remote-Sensing Techniques for Bridge Inspections

Visual-Semantic Alignments for Automated Interpretation of 3D Imagery Data of High-Pier Bridges