First Marine Geophysical Survey for Unexploded Ordnance Detection at Jikdo Military Firing Range

This study presents the first attempt to systematically detect and quantify unexploded ordnance (UXO) in the waters surrounding Jikdo, South Korea’s only live-fire maritime training range. A high-resolution marine geophysical survey was conducted using Side Scan Sonar (SSS) and a Multibeam Echo Sounder (MBES). To enhance detection accuracy and reduce subjectivity, a YOLO (You Only Look Once) deep learning algorithm was applied to acoustic imagery to identify UXO-like objects automatically. The detected targets were georeferenced using GIS and mapped onto bathymetric data to analyze spatial distribution. Results revealed that UXO-like anomalies were densely concentrated in the southwestern region of the survey area, particularly at depths between 40 and 50 meters—beyond the operational range of conventional diver-based methods. This study demonstrates the spatial limitations of traditional UXO detection and establishes the feasibility of integrating geophysical sensing with AI-based object detection for marine ordnance management. As the first application of this integrated method in Korean waters, the findings provide a foundational approach for data-driven, large-scale UXO risk assessment and clearance strategies in maritime military zones.