Crack detection in concrete elements from RGB pictures using modified line detection Kernels

摘要

Cracking is inevitable in ordinary reinforced concrete construction, yet it should be controlled to guarantee adequate serviceability and durability. Crack patterns analysis is essential for diagnosis, monitoring and maintenance, to identify malfunctioning or unsafe situations. An important part of this state is the cracking pattern of the elements as an indicator of the tensional distribution in the concrete. A method to search and measure fissures in RGB images of concrete elements has been developed and implemented as a MATLAB script. The algorithm uses several digital image processing tools to detect the measure, the cracks width and orientation; featuring these tools are the spatial filters called “orientation kernels” developed especially for detecting the angle and width of the cracks.