부산대학교 도서관

Remote surveillance of the ocean will soon become a high priority for the U.S. Navy, as international threats to close strategic choke points intensify, as piracy flourishes, and as gaps in U.S. waters continue to permit illegal intrusions with contraband cargo. A critical need is arising to identify threats as early and as distant from our shores as possible. A growing constellation of spectrally-capable satellites can facilitate this function, which must be performed autonomously. Earth's total ocean area is 1014 (1 m)2 pixels. This paper develops a spectral anomaly detection algorithm that is based on a statistical mixture model of clouds and ocean. A real time implementable prototype version is derived using clairvoyant fusion methods. Development of a second generation version applicable to a more accurate clutter model is also described.