摘要
基于改进三分量散射模型提出一种全极化合成孔径雷达(SAR)图像非监督分类方法。运用改进三分量分解模型解决体散射过高估计和负功率像素问题,提出类别重估步骤解决Wishart迭代聚类使聚类中心发生迁移的问题。首先对极化相干矩阵进行去定向操作,将各个像素的定向角旋转为0°;然后利用改进三分量分解模型将目标分解为平面散射、二次散射和体散射3种成分;接着利用3种散射功率计算功率散射熵,根据散射熵和3种散射成分的功率进行初步分类,利用Wishart迭代聚类优化分类结果;最后对Wishart聚类的结果进行类别重估,实现极化SAR图像的非监督分类。结果表明,本文算法的物理意义明确,分类结果易与实际地物相结合,测试区域的总体分类精度为98.6%,Kappa系数为0.973。
A polarimetric SAR image classification method based on the improved three-component scattering model[1]is proposed.First,the method performs a deorientation operation on the polarization coherent matrix,and the orientation angle of each pixel is rotated to 0 degrees.Then,the improved three-component scattering model is used to decompose the target into surface scattering,double-bounce scattering,and volume scattering.Then the power scatter entropy is calculated by using the three scattering powers.According to the scattering entropy and the power of the three scattering components,an initial classification is made.Then,using Wishart iterative clustering to optimize the classification results.Finally,the results of Wishart clustering are reassessed to achieve unsupervised classification of polarimetric SAR images.The results show that the algorithm has a clear physical meaning and the classification result is easy to be combined with the actual features.The overall classification accuracy is 98.6% and the Kappa coefficient is 0.973.
出处
《电子测量技术》
2017年第12期220-227,共8页
Electronic Measurement Technology
关键词
极化SAR
分类
极化分解
去定向
散射熵
PolSAR
classificatiom polarization decomposition
deorientation
scattering entropy
