RENet: Rectangular convolution pyramid and edge enhancement network for salient object detection of pavement cracks

Abstract

Automatic pavement crack detection is still a challenge due to road markings, rain traces, and diverse topologies. A rectangular convolution pyramid and edge enhancement network called RENet is proposed for accurate and robust pavement crack detection in this article. The rectangular convolution pyramid module is first built on deep layers so that the features can describe defects with different structures. The optimized contextual information and features of shallower layers are gradually merged into three resolutions. Subsequently, the hierarchical feature fusion refinement module and the boundary refinement module are applied to each branch. These two modules effectively promote the seamless fusion of features at various scales and make the model pay more attention to boundaries. Finally, the outputs of the three branches are integrated to obtain the final prediction map. Experiments conducted on two pavement crack datasets demonstrate that the proposed framework advances other state-of-the-art algorithms in terms of robustness and universality.