Cracks in concrete structures should be measured periodically to assess potential problems in durability and serviceability. Conventional crack measure- ment systems depend on visual inspections and manual measurements of the crack features such as width, length, and direction using microscope and crack gage. However, conventional methods take time as well as manpower, and lack quantitative objectivity resulted by inspector. In addition, these have difficulties in measuring inaccessible surface cracks.
In this study, a measuring and analyzing system for concrete surface cracks is developed by employing a CCD(Charge Coupled Device) camera in combination with image processing and artifical neural network. This system consists of three major parts: (1) crack extraction, which can easily detect fine cracks using improved algorithm on the basis of binarization and shape analysis, (2) crack analysis, which is mainly focussed on calculating width, length, and direction of extracted crack image, and (3) pattern recognition, which is able to classify cracks into five types including horizontal, vertical, -45°-diagonal, +45°-diagonal, and random cracks using MLP(Multi- Layer Perceptron) model.
To examine validity of the system developed in this study, crack analyzing tests are performed on the images obtained from various types of concrete surface cracks. The test results revealed that the system is highly effective in automatically analyzing concrete surface cracks in terms of features and patterns of cracks.