Support-vector-machine-based method for automated steel bridge rust assessment

Computerized methods have been used for structure health monitoring and defect recognition in the civil engineering field for many years. However, there are still non-uniform illumination problems that require more research efforts to resolve.In view of this, a new support-vector-machine-based rust assessment approach (SVMRA) is developed in this research for steel bridge rust recognition. SVMRA combines Fourier transform and support vector machine to provide an effective method for non-uniformly illuminated rust image recognition. After comparison with the popular simplified K-means algorithm (SKMA) and BE-ANFIS, it is shown that the proposed SVMRA performs more effectively in dealing with non-uniform illumination and rust images of red- and brown-color background over SKMA and BE-ANFIS.

► SVMRA combines support vector machine and Fourier transform.
► SVMRA can handle non-uniform illumination problems for rust image recognition.
► SVMRA can effectively recognize rust on images with red or brown background colors.
► SVMRA outperforms SKMA and BE-ANFIS methods.