Robust surface crack detection with structure line guidance

Crack detection plays a pivotal role in civil engineering applications, where vision-based methods find extensive use. In practice, crack images are sourced from Unmanned Aerial Vehicles (UAV) and handheld photography, and the balance between the utilization of global and local information is the key to detecting cracks from images of different sources: the former tends to eliminate interferences with a global perspective, whereas the latter pays more attention to the description of local details of cracks. However, many existing methods primarily target crack detection in handheld photographs and may not perform optimally on UAV-generated images or those with variable backgrounds or from different sources. In response to this challenge, we propose a robust and innovative method called Crack Detection with Structure Line (CDSL). The primary steps of this method can be summarized as follows: first, based on local information, an indicator called the “crack measure” is derived to directly generate a continuous crack map for effective image binarization; then, based on global information, the crack map is simplified in a unified and analyzable form using structure lines to perform a robust optimization for high-precision crack detection. The experiments we conducted on two publicly available datasets showed that CDSL provided competitive crack detection performance and outperformed four classical or current state-of-the-art methods by at least 13.0 % in the UAV dataset we collected.