Visual recognition and assessment of concrete crack properties

Post-earthquake structural safety is evaluated manually by certified construction inspectors and structural engineers. They inspect visible damage (e.g. cracks) on critical structural elements (e.g. concrete columns). This process is time-consuming and costly. In order to automate this type of visual inspection, several damage detection methods have been developed. However, most of them can only detect the presence of damage. The next step, retrieving damage properties and further using these properties to assess the impact of the damage on the structural elements has not yet been adequately investigated. This paper presents a novel method of retrieving crack properties. Under this method, the crack map on a structural element surface is first produced using state-of-the-art crack detection techniques. Then, the topological skeletons of cracks are created through binary image thinning, and the distance field of crack pixels in the map is produced using a distance transform. According to skeleton configurations and the distance values of the crack pixels, crack properties (width, length, orientation and location) are calculated. These properties are related to the properties of the structural element to produce relative measurements for the estimate of concrete column damage states. The method is implemented in a C++ based prototype and tested on a set of real crack images.