Concrete Crack Detection using Relative Standard Deviation for Image Thresholding

One of the biggest challenges with the reinforced concrete structures and concrete buildings are to identify the possible cracks. Cost-effective automated structural maintenance systems have attracted considerable attention in the recent past to detect the cracks in concrete structures. In this paper, we propose a concrete crack detection system based on image processing technique. For pre-processing an image, the anisotropic diffusion process is used and the top-hat and bottom-hat filtering techniques are used to enhance the contrast of an image. Top-hat filter enhances bright objects of interest from a dark background and the bottom-hat filter does the opposite. The existing algorithms require manual calibration of the threshold of some predefined specific parameters. However, we are using a new strategy for threshold selection based on relative standard deviation; this makes a difference in the image segmentation. The proposed method is tested with almost 200 images from public datasets and the experimental results prove that this method is efficient and effective to detect cracks in the concrete.