Unsupervised Method For Sea-land Segmentation And Oil Spill Extraction Using SAR Image

ABSTRACT:Oil spill not only affects the development of economy, but also damages the circumstance of the ocean, and brings great disaster to marine animals and plants and the people living along the coast. It’s important to build up a real-time and accurate marine monitor system. This paper focus on the real-time and non-supervised monitor system points, choosing the low-cost, full-time and full-weather SAR (Synthetic Aperture Radar) to build up the system. The study consists of non-supervised sea-land segmentation and non-supervised oil spill extraction, and experiments were done to prove the feasibility. And a system framework for Oil Spill Detection is built.The main contents are as follow:1) Brief introduce the theory of SAR and the features of SAR image, then introduce the background and significance of the study, along with the clue of this paper from some existing methods.2) First, introduce the common methods for segmentation. According to the level set of C-V model, this pager gives the improvements to realize the non-supervised sea-land segmentation. And the experiments are implemented to prove the improved method.3) In order to accomplish non-supervised method for oil spill extraction, this paper introduce a method convert the segmentation to classification, and with the help of CFAR to extract the dark area of each classification. At last, we build up a system of Oil Spill Detection.4) The conclusion and what should be done in the future are given.