Automatic crack inspection for concrete bridge bottom surfaces based on machine vision

In bridge buildings, concrete is widely used because its materials are considerably low-cost and it has high plasticity. However, some drawbacks exist in this kind of bridges, and crack is the most common ones. In order to avoid the cracks in bridge buildings becoming worse, it is necessary to periodically perform the inspection for it. Thus, a bridge inspection robot system with machine vision is designed for precise and robust bridge crack detection. In order to facilitate the analysis for cracks, a number of images are collected and are stitched into a high quality panorama, then the crack-like defects in the panorama are segmented. Firstly, in this paper, a quick and high-quality method for image stitching is applied, which is based on ORB algorithm. Then, the local directional evidence(LDE) method is used to enhance the crack structures from low contrast images, which serves as a preprocessing. Finally, the crack-like defects can be easily segmented by several morphological operations and a technique called Tubularity flow field. The experimental results have not only verified the rapidity and high-quality of applied image stitching method, but also the excellent effect of the segmentation method.