A Large-Scale Benchmark for Pavement Crack Segmentation

Pavement crack segmentation based on images is important for detecting and repairing pavement cracks in a timely manner. In this paper, we collect a dataset consisting of 7000 pavement crack images, called PCSeg7000, to reflect real road scenarios. Then, we propose a pavement crack segmentation method based on the U-Net framework. Our method combines a strip-attention mechanism and a self-supervised pyramid network structure to the U-Net framework, enhancing crack feature extraction and fusion results. We report experiments on the PCSeg7000 dataset. Compared to the original U-Net method, our method achieves higher accuracy, recall, and F1-measure, with improvements of 1.60%, 0.47%, and 1.04%, respectively. Additionally, our method outperforms some existing methods on other public datasets.