Abstract:Satellite remote sensing-based shoreline extraction can provide great supports for scientific planning and utilization of coastal zones, marine ecological protection, and efficient shoreline management. In this paper, based on the designed dual-loop network framework, we propose an adaptive image segmentation algorithm to achieve end-to-end shoreline extraction. Specifically, by considering triple constraints, a label reassignment module is firstly designed to enabling the extraction network with self-supervised learning ability for adaptive segmentation. Secondly, a data stream attention scheme is constructed to guide the networks learning directionally, which could enhance the robustness and convergence speed of the dual-loop network. Finally, a series of experiments are carried out to validate the effectiveness and feasibility of our proposed work. The dataset is available at https://pan.baidu.com/s/11qQvt7yuvJFTgfyWu5XRnQ?pwd=dy2y . Furthermore, the algorithm achieves superior performance, with a 15.5% improvement in accuracy compared to the latest algorithms.

A shoreline extraction method based on dual-loop network framework