Image analysis for the automated study of microcracks in concrete

An image analysis technique for the automated detection and quantification (at the mesoscopic scale) of microcracks in cement-based materials is described. In order to highlight microcracks and other microdefects, the surface of the material is impregnated with a red or a fluorescent dye. Observations of the polished sample are carried out using an optical microscope. The image analysis algorithm developed for the segmentation of the microcracks includes a pretreatment of the color image, an automatic thresholding carried out on the gray level histogram and a treatment of the binary image. The resulting image is then treated using a shape analysis and edge corrections. The characteristics of the crack network are quantitatively determined using classical stereological methods. The technique enables the user to draw maps of the sample crack network over an area of several tens of square centimeters. Results obtained for undamaged and mechanically loaded samples are presented. (C) 2001 Elsevier Science Ltd. All rights reserved.