基于新极化特征参数的SAR海洋溢油检测

为了提升海上油膜与其他目标的可分离程度，提出基于特征值分解的一种新的极化特征G，该特征不仅能够反映集合中不同目标之间的极化状态，还能够描述不同散射类型在统计意义上的不纯度。若某个区域中去极化状态越弱，不纯度越低，则该区域中新极化特征G的值越低。利用两景Radarsat-2全极化SAR (Synthetic Aperture Radar)影像对新特征的有效性进行实验验证。结果表明：海水具有较小的特征值，油膜具有较大的特征值，生物膜的特征值介于两者之间。且与span、\bar{\mathrm{\alpha }}、P、A、CPD等5种经典的极化特征相比，新特征在图像对比度、局部标准偏差及概率密度曲线等三个指标上均有更好的表现，不仅能区分生物膜（植物油模拟）与原油，且具有更好的抑噪性。

New Polarimetric Feature Parameter for Marine Oil Spill Detection in SAR Images

In order to improve the separability of oil film and other targets, a new polarization feature G based on eigenvalue and eigenvector decomposition is proposed. The new feature can not only reflect the polarization states between different targets in the corresponding set, but also has the ability to describe the statistical information impurities of the different scattering types. If the depolarization state was weaker, the impurities were smaller, then the value of the new polarimetric feature G in the specific region would be lower. Two sets of Radarsat-2 fully Pol-SAR (Polarimetric Synthetic Aperture Radar) data are used to verify the validity of the new feature G. The results show that there is a small eigenvalue in the seawater, a large eigenvalue in the oil film, the eigenvalue of the biofilm is between the oil film and seawater. In addition, the new feature G have better performance than span, \bar{\mathrm{\alpha }}, P, A and CPD in the image contrast, local standard deviation and probability density curve, which proves that the new feature G not only can distinguish bio-film(simulated by plant oil) and crude oil, but also has a good noise suppression ability.