Enhancing drowning detection in coastal environments via polarization image fusion

Drowning accidents occur frequently in coastal bathing areas; however, the complex natural environment presents substantial challenges for automated detection. In particular, the presence of specular reflections on the sea surface and the interference caused by water disturbances pose significant challenges to conventional visible-light-based methods, leading to substantial reductions in detection accuracy. To overcome these limitations, this paper introduces polarization imaging and leverages the complementary characteristics of different polarization modalities to enhance the visibility of drowning targets. Specifically, the polarization intensity image (S0) effectively suppresses sea-surface reflections, thereby mitigating background interference, while the degree of linear polarization (DoLP) significantly improves the contrast between swimmers and the water surface, making them more distinguishable under complex conditions. Building upon these advantages, we propose a multimodal feature fusion framework for drowning detection, in which a hierarchical fusion strategy integrates S0 and DoLP to generate fused images that simultaneously achieve background suppression and target enhancement, thereby improving the robustness of detection models in real-world scenarios. To validate the effectiveness of the proposed method, drowning-related image data were collected from a real coastal bathing environment and systematically evaluated. Experimental results demonstrate that the proposed method achieves superior qualitative and quantitative performance across diverse scenarios, effectively enhancing drowning detection accuracy. To the best of our knowledge, this is the first study to introduce polarization image fusion into drowning detection, and the results verify its effectiveness in coastal bathing environments.