Structural Damage Detection and Quantification Using Image-Based Methods

Abstract

In this thesis, we developed methods using image processing techniques for automatic detection and quantification of cracks from 2D images of damaged structures, especially concrete structure. In the images, the cracks are treated as tree-like topology dark objects of which each tree branch is assumed to be line-like and have local symmetry across the crack center axis. Utilizing the geometric features of the cracks, we proposed a crack measure concept to enhance cracks and reduce non-crack objects in the images. Basing on the crack measure concept, we developed a novel crack measure-based B-spline level set model that can automatically extract the level set function of cracks from the 2D crack image with intensity inhomogeneity. A new cost functional together with a crack measure technique is introduced to derive an iterative procedure for obtaining the exact level set function of the crack image via an optimization approach. From the converged level set, which is a binary image, we proposed another method for crack quantification. In this quantification method, estimated crack centerlines are obtained by applying morphological thinning algorithm to the binary image of the converged level set. Then, the estimated center lines of the detected cracks are fitted by cubic splines and the pixel intensity profiles in the directions perpendicular to the splines are used to determine the edge points. The location of edge points are used to compute the width of cracks. Various experiments of real crack images are used to demonstrate the excellent performance of our techniques.

In this thesis, we developed methods using image processing techniques for automatic detection and quantification of cracks from 2D images of damaged structures, especially concrete structure. In the images, the cracks are treated as tree-like topology dark objects of which each tree branch is assumed to be line-like and have local symmetry across the crack center axis. Utilizing the geometric features of the cracks, we proposed a crack measure concept to enhance cracks and reduce non-crack objects in the images. Basing on the crack measure concept, we developed a novel crack measure-based B-spline level set model that can automatically extract the level set function of cracks from the 2D crack image with intensity inhomogeneity. A new cost functional together with a crack measure technique is introduced to derive an iterative procedure for obtaining the exact level set function of the crack image via an optimization approach. From the converged level set, which is a binary image, we proposed another method for crack quantification. In this quantification method, estimated crack centerlines are obtained by applying morphological thinning algorithm to the binary image of the converged level set. Then, the estimated center lines of the detected cracks are fitted by cubic splines and the pixel intensity profiles in the directions perpendicular to the splines are used to determine the edge points. The location of edge points are used to compute the width of cracks. Various experiments of real crack images are used to demonstrate the excellent performance of our techniques.