Performance evaluation of ship detection using the 250m Geostationary Ocean Color Imager-II (GOCI-II)

This study explores the preliminary results of ship detection using the 250-meter Geostationary Ocean Color Imager II (GOCI-II), which offers multiple daily observations. A shape-based algorithm was introduced to reduce false positives from oceanic features, clouds, or noise, improving an initial pipeline that uses spectral and morphological features. Detection rates reached about 77% for ships over 100 meters and 85% for those exceeding 200 meters. Performance varied with ship size, and seasonal changes, with a noted non-detection window between days 250 and 280. Nevertheless, we believe ship detection using geostationary satellites can help extend a new paradigm of satellite-based ship monitoring techniques.