Deep Learning-Based Sea Fog Detection by Using Region of Interest

초록

This paper proposes a novel deep learning-based sea fog detection scheme that leverages CCTV footage and predefined Regions of Interest (RoIs) to address the limitations of existing maritime monitoring systems. By segmenting the RoIs based on horizon lines and visibility distances, the proposed system effectively analyzes the stage-specific risk levels of sea fog. Through learning the visual patterns within each RoI, the proposed scheme accurately predicts the occurrence and density of sea fog under diverse maritime conditions. Experimental results demonstrate that the proposed scheme surpasses traditional CNN-based models across various performance metrics, including accuracy, precision, recall, and F1-score. Additionally, the proposed scheme achieves a fast processing of image frames, ensuring real-time applicability.

키워드