三维极化旋转域解译工具及舰船检测研究

Three-Dimensional Polarimetric Orientation Domain Interpretation Tool and Ship Detection

极化雷达有利于获取目标完整的极化信息,在许多应用领域得到了广泛应用。然而,目标相对于雷达视线的几何位置通常会对散射机制产生显著影响。近年来,极化旋转域解译方法成功地表征和利用了这种目标散射多样性。值得注意的是,极化响应同时受到极化方位角和极化椭圆率角的影响。对此,本文将极化旋转域解译方法沿极化椭圆率角维度进行拓展,旨在探索和利用更为完整的目标散射多样性。其主要思想是提出一种三维极化相关方向图,能够同时表征极化相关特征随极化方位角和极化椭圆率角的变化情况。通过研究揭示了不同典型散射体的三维极化相关方向图存在显著差异,着重表现在极值点的分布和曲面起伏程度的不同。在此基础上,本文引入了微分几何中的曲率来描述这种三维流形的性质,并提出了高斯曲率最大值这一新极化特征。结合星载极化合成孔径雷达(Synthetic Aperture Radar,SAR)实测数据验证了高斯曲率最大值特征能够有效地区分舰船和海杂波。结合该特征,本文提出了一种阈值分割的舰船检测方法。与传统的恒虚警率方法和新近提出的NPNF方法相比,本文方法表现出更好的舰船检测性能。尤其是在近岸舰船密集区域,检测性能得到明显提升。

Polarimetric radar can acquire the complete target polarization information and is widely used in many applications.However,the geometric relationship between target and radar line of sight usually has a significant effect on the scattering mechanism.Recently,the polarimetric rotation domain interpretation techniques have explored abundant information in such target scattering diversity.Note that,the polarization response is affected by both the polarimetric orientation angle(POA)and the polarimetric ellipticity angle(PEA).In this vein,this paper extends the polarimetric rotation domain to the PEA dimension.The main idea is to propose a three-dimensional polarimetric correlation pattern,which can characterize the polarimetric correlation values with the change of POA and PEA.Investigations show that the proposed polarimetric correlation patterns for different canonical scatterers are significantly different in terms of the extreme points’distribution and the degree of surface undulation.Thus,the curvature in differential geometry is introduced to describe the characteristic of this 3D manifold,and the Gaussian curvature maximum feature is proposed.The experiments with the polarimetric synthetic aperture radar(SAR)data demonstrate that the Gaussian curvature maximum feature can discriminate the ship targets and the sea area.Based on the proposed feature,a threshold segmentation method for ship detection is proposed.Compared with the constant false alarm rate method and the NPNF method,the method pre-sented shows better detection performance.Especially in the inshore dense area,the detection performance is significantly improved.