Maritime Target Detection Method Based on Feature Fusion of Visible and Infrared Images

To address the issue of low detection accuracy in single-modal maritime target detection, this paper proposes a Feature Level Fusion Method of Visible and Infrared Images based on Cross-modal Interaction (FLFM). The algorithm extends YOLOv8 into a dual-stream symmetric network architecture, incorporating a Cross-modal Interaction (CMI) module to effectively leverage complementary features from different modalities while mitigating noise introduced by direct feature concatenation. We further develop an improved Feature Fusion Module (FFM) that integrates Coordinate Attention (CA) mechanisms into the multi-scale feature fusion framework. This enhancement enables effective fusion of cross-scale features and contextual information while alleviating positional information deviation. Experimental results on maritime datasets demonstrate that compared with single-source detection, the FLFM framework significantly improves environmental adaptability in maritime detection scenarios and achieves substantial accuracy improvements.