Vision-based Crack Identification on the Concrete Slab Surface using Fuzzy Reasoning Rules and Self-Organizing

Abstrak

Identifying cracks on the surface of concrete slab structure is important for structure stability maintenance. In order to avoid subjective visual inspection, it is necessary to develop an automated identification and measuring system by vision based method. Although there have been some intelligent computerized inspection methods, they are sensitive to noise due to the brightness contrast and objects such as forms and joints of certain size often falsely classified as cracks. In this paper, we propose a new fuzzy logic based image processing method that extracts cracks from concrete slab structure including small cracks that were often neglected as noise. We extract candidate crack areas by applying fuzzy method with three color channel values of concrete slab structure. Then further refinement processes are performed with Self Organizing Map algorithm and density based noise removal process to obtain basic crack characteristic attributes for further analysis. Experimental result verifies that the proposed method is sufficiently identified cracks with various sizes with high accuracy (97.3%) among 1319 ground truth cracks from 30 images.