A CNN-Based Method for Concreate Crack Detection in Underwater Environments

Pre-stressed concrete cylinder pipe (PCCP) is used in pipelines for long-distance water conveyance. However, due to complicated underwater environments, many vision-based concrete crack detection methods cannot be directly applied to the internal surface of the PCCP. This research proposes a two-step method to automatically detect concrete cracks in underwater environments. The first step involves preprocessing of the images through illumination balancing and image smoothing. In the second step, the preprocessed images are sent to a convolutional neural network (CNN) for crack detection. This method can overcome such issues as uneven illumination, strong noise, and blurred images, and it can effectively detect and locate cracks in underwater images with low illumination, low signal-to-noise ratio, and low contrast.