Feasibility of Ship Detection and Tracking Using GOCI-II Images

Abstract

The availability of high-resolution synthetic aperture radar (SAR), optical, radio frequency (RF), and other types of satellite imagery has improved, accelerating their utility in maritime monitoring (e.g., ship detection and tracking). Although high-resolution images from polar orbit satellites are considered for ship detection, their application is limited by the long revisit cycles of the satellites. The geostationary ocean color imager (GOCI-II), a geostationary ocean color satellite of South Korea (resolution of 250 m), overcomes this limitation by observing the waters that surround Northeast Asia every 1 h (10 times a day). This study investigates the feasibility and effectiveness of GOCI-II satellite imagery for ship detection, based on the near-infrared band. We tested the possibility of ship trajectory monitoring using multitemporal GOCI-II data. The testbeds for the ship detection effectiveness of GOCI-II were Busan, Yeosu, and Jeju in South Korea. The results indicated a few false detections because the automatic identification system (AIS) time did not match the threshold-section setting. However, several large and small ships were detected, with a major axis for each class. The detection rates for Busan (excluding the ships moored in ports) and Yeosu were 27 and 29.5%, respectively, confirmed through comparisons with the AIS data. Multitemporal ship detection and tracking was applied to a 50,000-ton ship navigating near Jeju Island, with an accuracy of 0–3 pixels. Furthermore, the possibility of detecting small ships (30-m class) was verified. This study can contribute to a paradigm shift in satellite-based ship monitoring by integrating high-resolution satellite imagery.