Ocean Oil Spill Detection Based On Fully Polarimetric SAR

The exploitation of offshore oil,leakage of ships,illegal discharge of pollutants,and leakage of natural oil have caused major losses to the marine ecosystem and natural resources.In order to effectively solve the problem of marine oil spills,remote sensing methods play a key role in oil spill detection.Synthetic aperture radar(SAR)has become an important mean for marine oil spill detection due to its all-weather and allweather advantage.In this paper,taking the ocean oil spill as the research object,polarimetric SAR data is used to detect the marine oil spill area.In the oil spill classification experiment,sevral classification methods were used to improve the oil spill classification accuracy.The main works in this article are summarized as follows:Oil spill feature from the extraction Polarization SAR: in this study,Radarsat-2 full-polarization data was used.The features of polarization total power,polarization decomposition,consistency parameters,and polarization degree were used to extract the features of the oil spill image.Comparative analysis of the extracted oil spill characteristics was implemented.Then J-M quantitative method was used to features selection,in order to obtain the optimal characteristics combination for oil spill detection.Improvement of Wavelet Neural Network Classifier Based on Genetic Algorithm: in order to detect the oil film from sea water,the selection of classifiers was particularly critical.As a classic three-layer network classifier,BP neural network was used for remote sensing image classification,but it did not work stably.Based on the BP neural network,the hidden layer basis function was replaced by the wavelet basis function,a wavelet neural network can be constructed with an improvement.However,the wavelet neural network classifier has a higher requirement for the initial value selection.In order to solve the initial value problem,the paper adopted the genetic algorithm to optimize the wavelet neural network.Finally,the effectiveness of the optimization method was verified by experiments.Multi-look polarimetric SAR classification method for the detection of oil spills: Due to the particularity of polarimetric SAR data,most methods reduce the 9 parameters of the polarization covariance matrix to one feature vector.In fact,the feature vector extraction was complex,and the statistical characteristics of polarimetric SAR data or physical scattering mechanisms can be used for image classification.Wishart classification method was used in the paper to classify the oil film from sea water.