Attention-Based Multi-Feature Fusion Method for Small Target Detection on the Sea Surface

This paper presents an Attention-Based MultiFeature Fusion Detector (AMFFD) to address the challenge of detecting weak and small targets in complex sea clutter environments. The proposed method fuses radar echo data represented by range profiles and range-Doppler maps, capturing both temporal and frequency domain information. A multi-level feature extraction module is employed to obtain rich representations, followed by an attention mechanism that adaptively enhances salient features while suppressing background interference. The enhanced features are further refined through feature aggregation and dimensionality reduction to extract critical information for target discrimination. Experimental results on the IPIX dataset and a real-world maritime radar dataset demonstrate that AMFFD significantly improves detection accuracy and reduces false alarm rates compared to existing approaches, highlighting its effectiveness in real-world sea surface surveillance applications.