Study On Identification Of Cracks From Straddle Monorail Track Beam Surfaces

Straddle monorail is one of the main forms of urban rail transport systems. The track beams of the girder bridge subsystem of the straddle monorail are both the bearing structures and the running tracks of the monorail train, so their health status is very crucial to the life safety of the passengers in the train and the pedestrians on the ground. The track beams are also prestressed concrete structures, which will suffer from the fatigue damage caused by the load and other factors. And thus, surface crack detection is an essential task in the daily maintenance of the straddle monorail. Benefiting from the rapid developments of digital imaging technologies, the existing crack image acquiring systems are becoming mature. However, the image processing technologies for crack detection remain relatively backward. Therefore, this paper is devoted to developing automatic crack identification methods based on image analysis, and focuses on solving the problem of identifying small cracks from noisy background. The computational speed is also taken into account at the same time.To identify the cracks as early as possible from the massive image data, the computational speed must be taken into account. Therefore, the first method, i.e., surface shape recognition method, proposed in this paper aims at classifying the massive images into crack and non-crack images at a relatively fast speed, so as to reduce the workload of the following crack extraction procedure, which is usually more time consuming. In this method, we first convert the images into parametric surfaces and then detect the crack surfaces using shape recognition techniques. The shape variations among non-crack surfaces are caused by approximately isometric deformations, but the dissimilarities between crack and non-crack surfaces are produced by non-isometric deformations. Therefore, the two classes of surfaces can be discriminated by their geodesic distance maps. To tackle the disturbances caused by the gaps in cracks, we develop a dedicated method, steady marching method, for the computation of the distance maps. This method is based on global geometric properties and thus very robust to noise. In addition, it can be extended for crack extraction.Since the severity evaluation of a crack is dependent on the extracted crack from image, crack extraction is a very important procedure of crack image analysis. However, it is still very challenging to extract the small cracks embedded in noisy background. Even some very recent methods require manual intervention or omission of crack width. In this paper, we aim at extracting such inconspicuous cracks automatically with width information preserved and propose two solutions.The basic idea of the first crack extraction method is to assign the pixel points to some arbitrarily shaped clusters, and then sift out the crack clusters according to their elongated shapes. Treating each gray-level image as a parametric surface, we devise an anisotropic clustering criterion that exploits the geometric properties of the surface. By virtue of the geometric representation and the anisotropy, this algorithm solves the problem of separating adjacent objects while simultaneously grouping the fragments of a crack into the same cluster. Moreover, the globally convex segmentation model is incorporated into our method, serving as a tool that provides appropriate candidate points and important parameters for the clustering procedure.The second crack extraction method is named as minimal path point process. The key problem faced by the traditional minimal path-based methods is how to set the endpoints of a minimal path automatically. In our method, we treat the two endpoints of a minimal path as a marked point of the point process, and then establish the Gibbs energy of the point process. In this manner, the problem of automatic endpoint location can be solved by minimizing the Gibbs energy. In contrast to the exsiting marked point process methods, the proposed model dispenses with the regularization term, and the configuration space is much smaller. Therefore, the optimal endpoints can be found efficiently by the stochastic searching algorithms. By integrating with the globally convex segmentation model, this method is also able to obtain the crack width information. Compared with the first crack extraction method, the minimal path point process is more robust to the cracks with very poor continuities or vague boundaries.The final method proposed in this paper is used for local shape analysis of the crack extraction results, so as to locate the widest crack sections and remove the lump noises. We continue to adopt the geometric representation of images, and then choose the Laplace-Beltrami eigenfunctions as the analysis tool of local shape to locate the widest crack sections and the lump noises. According to the characteristics of crack surfaces, we also modify the Laplace-Beltrami matrix to make it work on these special surfaces.