ROBUST IMAGE PROCESSING FOR DETECTION OF CONCRETE CRACKS USING A PARALLEL IMAGE-FILTER

Abstract

This paper presents a new image processing for stably detecting cracks from various surface-images of concrete structures and identifying width and direction of them. The process for detection consists of two steps. The first step is to build a robust image-filter for detecting major cracks with the aid of genetic programming. The second step is to remove noises and to trace obscure cracks by iterative applications of the image-filter in local regions. The proposed method was found to stably detect cracks from photos taken under the various conditions. Then, widths and directions of the detected cracks are identified on the basis of the spatial derivatives of brightness patterns, and it is found that the identified values show good agreements with those by visual inspection