基于半正定约束的极化相似度最优模型匹配目标分解

Positive-semidefinite Based Target Decomposition Using Optimal Model-matching with Polarization Similarity

目标分解是实现极化合成孔径雷达目标分类、检测与识别应用的重要手段。传统方法由于优先对体散射分量进行提取,其体散射能量的高估或二面角散射能量的低估现象较为严重。该文通过引入极化相似度量,基于数据驱动自适应地对基本散射机制的最优匹配模型进行选择。在此基础上,根据极化相似度量确定基本散射机制散射能量提取的优先顺序,并以各阶次剩余矩阵能量非负为约束,最终确定面散射、二面角散射、体散射这3种基本散射机制的能量贡献值。实测数据处理结果及其与光学图像的对比结果表明,该文方法获取的极化目标分解结果优于传统方法,能够准确地提取目标区域的基本散射特征。

Target decomposition is an important tool to realize target classification, detection and recognition applications with Polarimetric SAR （PolSAR）. However, the traditional method with priority of volume scattering component extraction seriously performs overestimation in the volume scattering energy or underestimation in the dihedral scattering energy. In this paper, by introducing polarimetric similarity measure, data-driven modelmatching for basic scattering mechanism is proposed. On this basis, the priority of scattering mechanisms energy extraction is determined with the similarity measure. Based on the non-negative constraint of energy, all the orders of residual matrix are reextracted for the final energy contribution of the dihedral scattering, volume scattering, and surface scattering mechanism. The processing results of real data and their comparison with the optical image results show that the proposal is better than traditional methods for the accurate extracttion of the basic scattering characteristics in the targets region.