基于SERD与SPAN的全极化SAR数据海岸线分类

发布时间：2018-04-30 15:12

本文选题：极化SAR分类 + H/α-Wishart分类器　；参考：《海洋测绘》2016年02期

【摘要】：基于Wishart分类器的全极化SAR图像H/α分类方法应用于海岸线分类,可区分不同海岸线类型。在聚类过程中,传统H/α-Wishart方法时常将各类聚类中心对应的散射机理混淆,使得同为单次面散射的淤泥质与砂质海岸线类别难以区分。针对此问题,提出了改进型的辅助分层分类方法。算法运用单次反射特征值相对差异度与极化总功率系数构成的二维特征空间,结合支持向量机得到最优分界判据,以分离初始分类结果中混淆的淤泥质与砂质海岸线。实验表明,改进方法能够有效区分淤泥质与砂质岸线,分类混淆矩阵对应kappa系数由0.794提升至0.853,分类识别率得到显著提高。
[Abstract]:The full polarization SAR image H / 伪 classification method based on Wishart classifier is applied to shoreline classification, which can distinguish different shoreline types. In the process of clustering, the traditional H / 伪 -Wishart method often confuses the scattering mechanism corresponding to various clustering centers, which makes it difficult to distinguish the silt from the sandy shoreline. In order to solve this problem, an improved hierarchical classification method is proposed. The algorithm uses the two-dimensional eigenspace composed of the relative difference of the single reflection eigenvalue and the total power coefficient of polarization, and combines the support vector machine to obtain the optimal boundary criterion to separate the muddy and sandy shorelines confused in the initial classification results. The experimental results show that the improved method can effectively distinguish muddy and sandy shoreline, and the corresponding kappa coefficient of classification confusion matrix is increased from 0.794 to 0.853, and the classification recognition rate is improved significantly.
【作者单位】：湖南省第一测绘院;中国科学院遥感与数字地球研究所遥感科学国家重点实验室;中国科学院大学;
【基金】：国家自然科学基金(41371355) 国家自然科学海外及港澳学者合作研究基金(41228007) 2015年遥感与数字地球研究所研究生所长基金资助课题
【分类号】：P229

【相似文献】