Research On Extracting Information Of Ocean Oil Spill Based On SAR Image

The ocean is the largest ecological environment in the world, and it is a crucial role for regulating global climate. Meanwhile, it is a huge treasure of mankind because of its rich biological and mineral resource. As active increasingly human activity in the ocean, it is gradually being put on the agenda to protect marine environmental. It is the most important protective measure to treat oil spill in the ocean. In order to treatment oil spills at sea timely, the oil spill region should be detected as early as possible. Traditional optical remote sensing, due to the impact of rain, snow and the night, receives a lot of restrictions. Synthetic Aperture Radar(SAR) can penetrate clouds and work in the dark, so it is an all-weather and all-time radar. AS SAR and its theory become more and more advanced, it receives more attention to detect oil spills at sea in microwave remote sensing.This paper analyzes the SAR imaging theory and polarization decomposition theory. O n this basis, the algorithms of extracting information about oil spill in the field of quadruple and single polarization respectively, and the main work is as follows:Distinguish oil-alike and oil spill in single-polarization field: 2D-Otsu algorithm is used in image segmentation. Before image segmentation, the universal processing has been achieved to improve the application scope of the algorithm. After image segmentation, a number of morphological filtering is carried out to enhance the quality of segmentation. N ine feature parameters including complexity, the edge gradient, etc., are selected, and they are considered as evidences to determine whether the aim is oil spill. Based on BP neural network in pattern recognition, a classification judger about oil spills is established, and the characteristic parameters extracted from the set of sample image are used to train the neural network. Then, the neural network is used to verify the algorithm of the whole process, and a satisfactory result is achieved.Obtain oil spill information in quad-polarization field: Firstly, we analyze all phase differences among many types of at sea target associated with oil spills in various polarization field, and it is indicated that co-polarization phase difference is effective to identify biological oil slick which belongs to oil-alike. Based on this theory, ? filter, which relates to the standard difference of co-polarization phase difference, is built. Then,the phase difference images are filtered with ? filter to identify biological oil slick. H- ? images of sorts of aims related to identifying oil spill study, and the reason why the traditional H/ ? is often not a good arithmetic in recognition oil spill field is exposed. The anti-entropy in C loud decomposition is introduced, and the Wishart clustering method is applied in oil spill identification also. The H/A/ ?-Wishart arithmetic is used to identify oil spill in ocean, and the classification accuracy is improved. Lastly, the real SAR images are used to verify H/A/?-Wishart algorithm, and the presence of a problem is indicated.