Improving Accuracy in Detecting Unknown Objects and Enhancing Low Visibility Conditions Caused by Sea Fog in Coastal Areas

South Korea, being in a tense situation with the North and surrounded by the sea on three sides due to its geographic characteristics, places great importance on coastal surveillance for national security. However, challenges arise in coastal surveillance operations due to outdated military equipment and a reduction in military personnel caused by low birth rates. Against this backdrop, this paper presents deep learning based automation technology as a substitute for human resources. Coastal regions often experience low visibility due to sea fog caused by unique climate conditions. To address this issue, the Dehazy algorithm is introduced for fog removal, and an algorithm for background separation is implemented by dividing the boundary between the sea, sky, and obstacles based on the horizon to focus on objects on the sea. For object detection, the YOLO algorithm is used, and this paper highlights the difference in object recognition rates and real time processing speed when identifying unidentified objects, both in the original and processed images.