Bridge Crack Detection Based on Image Segmentation

Abstract

The detection of bridge cracks is related to the life of the bridge. Manual detection is time-consuming and laborious. Contact sensors exposed to air are susceptible to weather damage. In practice, the bridge cracks with low contrast and blurred edge features is the difficulty of crack detection based on image segmentation. To this end, this paper proposes a deep learning based image segmentation detection network. In order to reduce the size of the network model, we modify the backbone network of Segnet. The feature extraction network is modified to the structure of mobilenet and improved. Cracks belong to small targets and easily missed in the detection process. In order to improve the detection accuracy of small targets, a multi-scale feature fusion operation is adopted in this paper. The network training uses public datasets. In some images, the contrast between the crack and the background is low, so this paper binarization is used to strengthen the crack structure. The experimental results verify the effectiveness of image segmentation.