Nearshore Ship Detection in Optical Remote Sensing Images Using Feature Decoupling and Instance Segmentation
Aiming at the problem of inshore ship target detection in optical remote sensing images,a deep learning ship detec-tion method based on case segmentation and feature decoupling is proposed.Adaptive feature pyramid is used to automatically learn and fuse multi-scale context to enhance the ability of feature extraction.The boundary feature disentanglement network is adopted,and the boundary prior knowledge is introduced to reduce the missed detection problem caused by the close arrangement of ships and the false alarm problem caused by the similarity between ports and ships,and improve the network detection performance.The center point prediction is used to further segment the network results,so as to further improve the accuracy of ship detection.A re-mote sensing near-shore ship detection data set is built.Experiments on the self-built remote sensing near-shore ship detection da-taset show that the method in this paper is superior to the traditional anchor method and has better detection performance.Thesis al-gorithm effectively improves the accuracy of near-shore ship target detection,and has reference value for nearshore ship detection under clutter background.
万方数据
维普
