논문 상세보기
Global Trend of Anomaly Detection From Satellites for Nuclear Verification –An Analysis of Publications
As remote sensing measures, satellite imagery has played an essential role in verifying nuclear activities for decades. Starting with the first artificial satellite, Sputnik 1, in 1957, thousands of satellites are currently missioning in space. Since the 2000s, the level of detail in pixels of an image (spatial resolution) has been significantly improving, thereby identifying objects less than one meter, even tens of centimetres. The more things are identifiable, the wider regions become targets for observation. With the increasing number of satellites, computer vision technology is required to explore the applicability of algorithm-based automation. This paper aims to investigate the R&D publications worldwide from the 1990s to the present, which have tried to apply algorithms to verify any clandestine nuclear activities or detect anomalies at the site. The versatile open-source publications, including the IAEA, ESARDA, US-DOE national laboratories, and universities, are extensively reviewed from the perspective of nuclear nonproliferation (or counter-proliferation). Thus, target objects for applications are essentially located in nuclearrelated sites, and the source type of satellite sensors focuses on electro-optical images with high spatial resolution. The research trend over time by groups is discussed with limitations at the time in order to contemplate the role of algorithms in the field and to present recommendations on a way forward.
