A Shoreline Extraction Method Based on Double-Loop Network Framework

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 double-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 network’s learning directionally, which could enhance the robustness and convergence speed of the double-loop network. Finally, a series of experiments are carried out to validate the effectiveness and feasibility of our proposed work.