A Method For Crack Detection On A Concrete

Abstract

Recently interest in the automated crack detection of photographs of concrete structures for non-destructive inspection has increased. In general, there are a number of noises, such as irregularly lit conditions, shadows, blemishes and divots in the concrete pictures. These contribute to difficulties in the automatic detection of cracks. This paper introduces two pre-processing’s to eliminate such cracking noises. First, minor variations such as irregularly illuminated conditions and shading are eliminated from concrete images by subtraction pre-processing with a smoothed image. Second, a line philtre based on the Hessian matrix is used to highlight the line structures associated with cracks.  Finally, threshold processing is used to isolate cracks from the background. The output of the proposed method is evaluated by ROC analysis with 50 actual images. The experimental results show that the proposed method is successful in detecting cracks on loud concrete pictures.