Dual-Branch Neural Network for Sea Fog Detection in Geostationary Ocean Color Imager

Sea fog significantly threatens the safety of maritime activities. This paper develops a sea dataset (SFDD) and dual branch detection network (DB-SFNet). We investigate all observed events in Yellow Bohai (118.1{\deg}E-128.1{\deg}E, 29.5{\deg}N-43.8{\deg}N) from 2010 to 2020, collect images for each event Geostationary Ocean Color Imager (GOCI) comprise SFDD. The location image SFDD is accurately marked. proposed characterized by long-time span, large number samples, accurate labeling, that can substantially improve robustness various models. Furthermore, this proposes achieve holistic detection. poporsed DB-SFNet composed knowledge extraction module optional encoding decoding module. two modules jointly extracts discriminative features both visual statistical domain. Experiments show promising results with an F1-score 0.77 critical success index 0.63. Compared existing advanced deep learning networks, superior performance stability, particularly mixed cloud areas.