Patch-based Crack Detection in Black Box Road Images Using Deep Learning

This paper proposes a method for patch-based crack detection of black box road images, for efficient road pavement monitoring. The proposed method is based on deep learning and consists of two modules: road extraction and crack detection. The road extraction module uses the segmentation process of a Fully Convolutional Network (FCN) called FCN-8s to leave only the road area in the image. The crack detection module performs patch-based crack detection on the extracted road area using a convolutional neural network. To the best of the authors’ knowledge, the proposed method is the first attempt to detect road cracks of black box images, which are not orthogonal but skewed actual road images.