利用最小冗余最大相关和SVM的SAR图像海上溢油识别

Oil Spills Identification in SAR Image Based on mRMR and SVM Model

近年溢油事故频发,海洋污染日益严重。利用合成孔径雷达(SAR)卫星可以有效跟踪由海上溢油事故导致的油膜扩展情况。利用最大冗余最小相关-支持向量机(mRMR_SVM)算法进行SAR图像溢油识别,为溢油事故决策支持提供重要前提。首先采用mRMR提取最优特征向量集,对输入值进行降维;然后采用SVM算法解决油膜图像分类问题,同时选择径向基函数(RBF)为核函数;使用训练集训练该模型,调整模型参数;以测试集特征向量作为输入,利用训练好的模型进行溢油识别。实验结果表明,mRMR_SVM模型对SAR图像的油膜和类油膜识别有效,准确率为96.875%。

In recent years,oil spills have been frequent and marine pollution has become increasingly serious.Synthetic Aperture Radar（SAR） satellites can effectively track the oil slicks expansion caused by the oil spill.The Minimum Redundancy and Maximum Relevance-Support Vector Machine（mRMRSVM） algorithm is used to identify oil spills in SAR images and the recognition results provide important preconditions for oil spill accident decision support.First,the mRMR algorithm is applied to select the optimal eigenvector set,and the input values are reduced.Then the SVM algorithm is used to solve the oil spill image classification problem,and Radial Basis Function（RBF） is selected as the kernel function.The model is trained by using training sets and the model parameters are adjusted.The trained model is used to identify oil spills with the test set eigenvector as an input.The experimental results show that mRMRSVM model is effective for the identification of ＂oil slicks＂ and ＂look-alikes oil slicks＂ image,and the accuracy rate is 96.875%.