Research On Marine Oil Spill Detection And Analysis Based On Spaceborne Polarimetric SAR Image

With the rapid development of the global maritime industry,the oil spill accidents,such as oil-carrying ships collision,illegal sewage discharges,and pipeline rupture,occur frequently,increasing the risk of oil spill in the marine traffic environment.The rapid and effective identification and detection of oil spill are of vital significance for strengthening the emergency command of maritime traffic safety and oil spill accident,maritime search and rescue,and repairing the marine environment after disaster.Synthetic Aperture Radar(SAR)remote sensing technology breaks the limitations of traditional monitoring methods,and has been widely used in the field of marine oil spill detection due to its all-day and all-weather microwave imaging capability.With the continuous development and improvement of SAR system which gradually expands from single polarimetric to polarimetric SAR(Pol-SAR),the complex scattering matrices of images can be measured and more abundant "holographic" information can be obtained through Pol-SAR systems,which is conducive to analyze and interpret the scattering mechanism of oil slick,look-like and seawater,and then realize the extraction and detection of oil slick.However,there are still some difficult and challenging work in Pol-SAR oil spill detection,such as the study of difference in oil spill identification performance between Pol-SAR systems with different modes,the further expansion of polarimetric features space,and the insufficient utilization of polarimetric information.This paper conducts an in-depth research aiming at the existing problems in Pol-SAR oil spill detection,and analyzes and compares the recognition performance of scattering mechanism of oil spill under different mode polarimetric SAR based on the polarization theory.In the research,the dual-pol SAR with partial polarimetric information and the quad-pol SAR with complete polarimetric information were used to identify and detect marine oil slick.The following innovative work has been carried out:(1)Aiming at the calculation redundancy caused by the time dimension in the multi-temporal dual-pol SAR oil spill detection and the lack of comprehensive analysis of different oil-water boundary conditions,this paper firstly proposes an oil spill region of interest extraction algorithm based on the occurrence frequency of multi-temporal potential dark regions,which realizes rapid extraction of oil spill region in time series wide-range images,and effectively reduces the computation of spatial dimension.On this basis,by expanding the information abundance and comprehensively considering the utilization of the advantage features under different boundary conditions,an oil spill detection method based on the random forest module which takes into account the polarimetric-texture advantage features under different oil-water boundary complexity conditions is proposed,thus realizing the effective extraction of oil spills under different oil-water boundary conditions.Further,relying on the method proposed in this paper,taking the Sentinel-1A multi-temporal image under the same observation conditions in the central Caspian Sea as an example,this paper analyzes the spatial distribution and the annual inter-monthly dynamic variation trend of the oil spill.(2)Aiming at the expansion of polarimetric feature to quad-pol SAR oil spill detection,a new polarimetric feature parameter is proposed.In order to improve the discrimination between the oil slick,look-alikes and background seawater,and to further expand the polarimetric features space,this paper proposes a new set of combination polarimetric features based on the polarimetric scattering entropy H and improved anisotropy A12—H_A12.The results show that the oil spill detection capability of H_A12 combination is superior to that of traditional H_A combination,which can be used as an alternate strategy to the latter.Among them,the combination of H(1-A12)has the best performance,and the contribution of separability and classification are overall better than the other four types of polarimetric features,which can maintain superiority and robustness in different oil spill scenarios,highlight the oil slick information,and effectively suppress sea clutter and look-alikes information.(3)Aiming at the lack of construction,description,and full utilization of the polarimetric features spectrum in the current quad-pol SAR oil spill detection method,a novel oil spill detection method based on multi-dimensional polarimetric feature intensity model matching is proposed.This paper constructs a multi-dimensional polarimetric features intensity model of oil slick targets,introduces the spectral pan-similarity measure(SPM)to measure the similarity of feature spectrum between models and each samples,and then searches for and matches homogeneous unlabeled pixels,which implements the description and effective classification of the information of the relative thickness of oil slick under the polarimetric space.The proposed method produces better results than traditional similarity measures and classifiers by comprehensively considering accuracy and computational complexity.In this thesis,a systematic study of marine oil spill detection based on spaceborne polarimetric SAR images is conducted.Based on the scattering mechanism analysis of oil-covered waters,the oil spill detection methods under dual-pol SAR and quad-pol SAR systems are constructed respectively,and the experiment results verify the effectiveness and robustness of the proposed oil spill detection method through experiments.This research has important significance of ensuring maritime traffic safety and improving maritime emergency management capability,and lays down the theoretical foundation for the recognition and detection of marine oil spills in different scenes with polarimetric SAR images.