A Robust Ship Detection Method in Complex Sea Scenes of Spaceborne SAR Images Based on Density Maps

Abstract

Ship detection is an important research topic in the interpretation of spaceborne synthetic aperture radar (SAR) images, with wide-ranging applications in maritime traffic planning and ocean monitoring and management. In a vast sea area, various interferences such as clutter intensity variations and target scale changes pose significant challenges, resulting in considerable performance discrepancies in ship detection algorithms across different sea scenes. To address the issue of poor robustness in ship detection under the complex and dynamic conditions of sea scenes, a robust ship detection method in complex sea scenes of spaceborne SAR images based on density maps is proposed. Through modules such as scale adaptation and context awareness, the method enhances robustness to target scale variations and clutter environment changes in dynamic sea scenes. Utilizing ship density maps based on elliptical Gaussian kernels, it better captures the morphological characteristics of ship targets, effectively improving the generalization capability of the detection method in complex sea scenes. Experimental results on three different types of measured ship detection datasets of spaceborne SAR images show that the proposed method can achieve robust ship target detection.