Abstract
We developed a new method for crack width measurement with image analysis using multiscale LoG (Laplacian of Gaussian) kernel. This method makes it possible to measure the width of cracks by detecting the precise directions and the edge positions of them due to that the kernels can reduce noise while they can detect the structure of the objects. The measured width in pixels can be precisely transformed to the width in mm by using a calibration target which can be also used to correct the perspective in a photograph. The transformation errors can be also used as criteria for judging the image quality of photographs whether we can measure the widths with sufficient precision with them. In this report, we describe the algorithm of the method and show the experimental results at real tunnels. The experimental results suggest that our method can measure the width of cracks with sufficient precision for periodic maintenance operation of tunnels.
