Image Based Crack Identification in Real Concrete Surface Using Watershed Segmentation

Abstract

Crack detection is important for the inspection, diagnosis and maintenance of concrete building. However, conventional image based approaches cannot achieve precise detection since the image of concrete surface contains various types of noises like concrete blebs, stain, insufficient contrast and shading. Visual inspection and hammer sounding are the traditional tools used in concrete inspection. The condition rating assigned to a concrete surface is based largely on the results of visual inspection and on an estimate of defect size. Defects hidden from view typically are not picked up, especially defects that are more than a couple of inches from the surface. Additionally, this method is slow, qualitative and potentially hazardous for the inspector. The imaging approach proposed herein is to apply optical inspection techniques in a new way to develop accurate, global bridge inspection techniques. The proposed techniques are based on an optical inspection methods, namely: Watershed Segmentation. The watershed segmentation technique recognizes and measures defects and structural flaws by means of detection and visualization of intensity gradients on the surface of the target. Analysis of the segmented images provides a quantitative signature specific to various types of defects. Additionally noise reduction based on percolation model is proposed. The validity of the proposed technique has been evaluated by means of some experiments on concrete surfaces