SAR Ship Detection Based on Attention Pyramids Network for Complex Scenes
Lin Wang, Yuting Lü, Xiaochen Liu, Zaidao Wen · 2021 China Automation Congress (CAC) · 2021
As one of the core tasks of ocean monitoring, Synthetic Aperture Radar (SAR) ship detection can provide important clues for military intelligence acquisition and ocean exploration, which plays a vital role in both military and civilian fields. However, the current ship detection method in complex scenes always suffer the inaccurate feature extraction. In this paper, we proposed a new SAR ship detection method for complex scenes based on attention pyramids network(CSAPN). The CSAPN adopts the pyramid structure, and the convolutional block attention modules are applied into each process of multi-scale feature map extraction. In this way, the related features to objects will be enhanced and the unrelated features will be suppressed. The results on real Gaofen-3 SAR data demonstrate the algorithm proposed in this study have the capability to suppress the interferences in complex scenes with a excellent accuracy.