Assessing Stationary Maritime Traffic Density in Territorial Dispute Zones Using Open-Source Sentinel-2 Satellite Imagery

Abstract

Although Sentinel-2 images are seldom employed for object detection due to their lower resolution, their accessibility
and cost-effectiveness make them a valuable tool for continuous global maritime monitoring. This paper proposes
using publicly available satellite imagery to detect ships in the maritime conflict zones of the South China Sea,
aiming to assess the intensity of conflict through a fine-tuned YOLO object detection model. Ship density is
estimated by aggregating detected ships over each area, demonstrating that Sentinel-2 is a valuable data source
for this task. Additionally, the correlation between the number of ships detected in a given week and the number
of articles reporting on conflicts in the relevant maritime zones during the same time period was calculated.
Preliminary results suggest that this approach could serve as a valuable indicator for evaluating military tensions in
the region, enhancing conflict transparency and supporting risk reduction efforts.