SHIP DETECTION BASED ON FEATURE FUSION FOR DUAL-POL SAR IMAGES

Ship detection in synthetic aperture radar (SAR) data, utilizing CNN-based detectors, has yielded remarkable results across diverse domains. These detectors swiftly identify targets in intricate marine settings by extracting and comprehending sample features, including structural and textural elements. However, in the face of the issue of complex image quality, there is still room for improvement in their detection performance. To this end, we propose a ship detection method based on feature fusion for dual-pol SAR images. This approach involves the rational exploration of complementary information concealed within co-polarization and cross-polarization images through the designed joint feature mining, cross enhancement module, and adaptive gate fusion module. It enhances and fuses the polarization features and feeds them into the detection head for regression, effectively reducing false alarms and comprehensively improving detection performance. Experiments on dual-polarization SAR datasets demonstrate its effectiveness and state-of-the-art performance.