Automatic crack classification and segmentation on concrete bridge images using convolutional neural networks and hybrid image processing

Cracks are an indicator for a bridge’s structural health and functional failures. Crack detection is one of the major tasks to maintain the structure health and serviceability of a bridge. At present, the most commonly used crack detection technology is manual inspection, with the disadvantages of being highly labor-intensive and time-consuming. In this paper, a CNN-based (convolutional neural network, CNN) crack detection method is proposed. To automate quantitative measurements of identified crack, a hybrid image processing is proposed, as well. Firstly, a dataset is accumulated, including 12,000 cropped crack images and 19,500 cropped background images. Secondly, preprocessed images with the proposed method of Bilateral-Graying-Contrast (BGC) are fed into ResNet and VGG (Visual Geometry Group Network) for training and testing. Finally, automatic measurement system of bridge crack is developed, which is not prone to weakened shooting conditions. The results demonstrate that Resnet achieves accuracy of cracks to 97.44%, which is higher than VGG. Our crack measurement system significantly reduces the measurement error to 9.86% and can be assumed as a reliable method in the analysis of concrete bridge images.