基于权重散射特征的FSVM算法在极化SAR影像分类中的应用  

Classification of Full Polarimetric SAR Data Based on the FSVM Algorithm Considering Weighted Scattering Features

作　　者：柯宏霞[1] 刘国栋[1] 龚正娟 KE Hongxia;LIU Guodong;GONG Zhengjuan(College of Civil Engineering,Chongqing Jiaotong University,Chongqing 400074,China;Yunnan Provincial Map Hospital,Kunming 650034,China)

机构地区：[1]重庆交通大学土木工程学院,重庆400074 [2]云南省地图院,云南昆明650034

出　　处：《测绘与空间地理信息》2019年第4期17-20,共4页Geomatics & Spatial Information Technology

基　　金：重庆市前沿与应用基础研究计划项目(cstc2014jcyjA0915);重庆交通大学实验教学改革与研究基金项目(syj201405)资助

摘　　要：提出了一种基于权重散射特征的模糊支持向量机的极化SAR数据监督分类方法。首先,对极化SAR数据进行H/SPAN/A/α散射特征提取;其次,根据样本可分离度设置最佳散射特征权重参数C,得到最优分类输入数据(H/6SPAN/A/α);最后,利用FSVM算法对数据进行监督分类。实验结果证明,所提出的FSVMH/6SPAN/A/α分类结果优于FSVM-H/SPAN/A/α。A supervised method called Fuzzy support vector machine(FSVM)considering weighted scattering feature is presented.Firstly,the scattering features(H/SPAN/A/α)were extracted from the Pol SAR data.Secondly,depending on the sample separability,the optimum parameters of weighted scattering features were set to obtain the best input space(H/6SPAN/A/α)of the FSVM algorithm.Thirdly,the classification was performed by the FSVM algorithm.The classification results show that the proposed method has out-performed the FSVM-H/SPAN/A/αmethod.

关 键 词：极化SAR分类 样本可分离度 权重散射特征 模糊支持向量机 

分 类 号：P237[天文地球—摄影测量与遥感]