Fourier-Transform-Based Method for Automated Steel Bridge Coating Defect Recognition

Intelligent computerized system can analyze and process vast amounts of data instantaneously in the civil engineering domain recently. However, most of previously developed image assessment methods for steel bridge rust inspection could not directly deal with acquired images. The acquired images should be firstly classified into two groups, non-defect or defect, by naked eyes or using the method reported in Lee's PhD dissertation (Lee 2005). Lee's method works effectively with blue-coated steel bridge images in Indiana. However, its effect on other colors or in particular environmental conditions, such as non-uniform illumination, has not yet been explored. A defect recognition method for steel bridge surface was proposed by using Fourier Transform and color image processing method in order to adapt to various background colors and overcome the influences from particular environmental conditions, such as non-uniform illumination. Fourier Transform in this system is aimed to determine the existence of defects in an acquired image. The proposed method shows better performance than previous methods such as rust defect recognition method (RUDR) proposed by Lee in 2005.