STUDY ON CRACK DETECTION IN BRIDGE DECK PAVEMENT ON URBAN EXPRESSWAYS USING LINE-SCAN IMAGES

Abstract

 In periodic inspections of urban expressway bridge deck pavements, high-resolution line-scan images are used. The operator visually determines the presence or absence of cracks from the line-scan images. This process is time-consuming and labor-intensive. Therefore, existing research has devised a method to automatically determine the presence or absence of cracks on a segmented image basis. However, this method does not provide results that can replace the operator's work because it cannot identify the presence or absence of cracks on a pixel-by-pixel basis.

 In this study, we developed a method to automatically detect cracks in line-scan images pixel by pixel using image domain segmentation. In a demonstration experiment, cracks in dense-grained pavement and drainage pavement were detected with an F-score of approximately 0.8. Furthermore, the optimal image segmentation method for each pavement type, as well as effective image processing for crack detection were identified.