Intelligent real-time control system through socket communication using deep learning-based de-hazing and object detection in an embedded board environment

This paper is a paper on the development of an intelligent real-time marine signal light that applies Socket communication after sea fog remove and object detection using artificial intelligence and computer vision through camera video in an embedded board environment. The embedded board used NVIDIA Jetson Nano board, and for sea fog remove, deep learning-based FFA-Net with excellent sea fog and fog removal performance among current technologies was used. For human detection, YOLOv4 based on light-weight deep learning with excellent detection performance among current technologies was applied for the purpose of research. For learning and testing, customized data was built through RESIDE dataset, MSCOCO dataset and data labeling, and socket communication was used for information delivery.