Abstract:In the process of oil exploration, development, refinement, transportation and storage, due to accidents or operational errors, the leakage of crude oil or oil products from the reservoir will flow to the ground, water, and the beach or sea. As the different thickness of oil, the leakage forms a film, the phenomenon is called oil spill. Large-scale oil spills often has a serious impact on national ecology, society and economy. Based on the satellite remote sensing data, this paper introduces the change detection into the problem of marine oil spilling, comparing several change detection algorithms with each other, and accomplishes an oil spill detection system. Results show that, the change detection can effectively reduce the impact of geographical location and lighting conditions, and it has a stronger generalization ability than the classification methods trained on single image. The research also shows the new method can effectively separate the sea water region with light oil or heavy oil region.