结合纹理的支持向量机合成孔径雷达溢油监测

针对目前合成孔径雷达(Synthetic Aperture Radar,SAR)监测溢油存在的问题,在分类时考虑像元灰度的空间分布和结构特征;同时考虑分类时样本不足的缺陷,采用结合纹理的支持向量机(Support Vector Machine,SVM)遥感图像分类方法,进行溢油目标的识别。以发生在西班牙的"威望号"溢油事件为例,利用目标样本对以灰度共生矩阵法提取各种纹理特征进行了分析,指出均值、对比度、方差、熵和相异性能够较好地识别溢油目标。采用最小距离、最大似然和SVM分类器分别对溢油目标进行提取,结果表明SVM具有较好的分类精度。

Research on SAR Oil Spill Monitoring by SVM Based on Texture Features

In view of the limitations of Synthetic Aperture Radar(SAR) in monitoring oil spill,this paper proposes that the gray-degree spatial distribution and structure characteristic of pixels should be considered in the classification. Meantime,texture-involving Support Vector Machine(SVM) classification method in remote sensing image processing is suggested in identification of oil slick to solve the problem of insufficient sample numbers.The oil spill accident of ship Prestige happened in Spain is taken as the e...