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

摘要

The weathering steel having natural corrosion resistance was used more frequently as the structural components for steel bridges in Japan, showing a generally good behavior. However, in some cases, it is observed that some areas of them exhibit a very severe surface corrosion damage which causes the degradation of the structure. It is noted that the surface texture is an important indicator for the performance assessment of weathering steel. This paper presents an imaging-based method for condition rating of surface quality using wavelet transform (WT) and support vector machine (SVM). Initial images are characterized by image texture analysis utilizing two-dimensional wavelet decomposition, from which both local and normalized global energy distribution of each detail sub-image are extracted as the feature vector. The SVM is then employed to classify a set of texture features into various corrosion levels. The effectiveness of the integrated method has been confirmed through systematic investigation on a representative set of 530 images.