基于加权投票集成的极化SAR图像分类方法  

PolSAR image classification method based on weighted majority vote ensemble

作　　者：陈博[1,2] 王爽[1,2] 焦李成[1,2] 

机构地区：[1]西安电子科技大学智能感知与图像理解教育部重点实验室,陕西西安710071 [2]西安电子科技大学智能感知与计算国际联合研究中心,陕西西安710071

出　　处：《华中科技大学学报（自然科学版）》2015年第3期79-82,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家重点基础研究发展计划资助项目(2013CB329402);国家自然科学基金资助项目(61271302;61272282;61202176;61271298);高等学校博士学科点专项科研基金资助项目(20100203120005)

摘　　要：提出了一种新的基于加权投票准则集成的极化合成孔径雷达系统(PolSAR)图像分类方法.该方法采用加权投票集成的方法根据不同个体的学习结果进行合,并从而提高极化SAR图像的分类精度.首先,输入极化图像数据并获得所需要的特征作为特征集.再从图像的每一类中选取多组像素点组成多个训练样本子集;然后,基于不同的样本子集训练学习得到不同的分类器,并对像素点进行分类得到预测标记,再由这些预测标记计算得出相应的加权系数;最后,通过加权系数将预测标记合并起来得到最终的极化SAR分类结果.实验结果证明,所提出的算法在AIRSAR和Radarsat-2数据上取得了很好的分类结果.A polarimetric synthetic aperture radar (PolSAR ) image classification method based on weighted majority vote ensemble was proposed .The weighted majority vote ensemble was adopted to learn on different training samples ,in order to improve the classification results .Firstly ,the features were extracted from the PolSAR data ,and several groups of pixels in one class were chosen as the training sample subsets .After that ,the component classifiers learn on different training samples gave the predictive labels for the pixels ,and the weights were calculated on these labels .Finally ,the pre‐dictive labels were combined together to get the final classification result .The experimental results demonstrate the effectiveness of the proposed method on AIRSAR and Radarsat‐2 data .

关 键 词：图像分类 雷达极化 监督分类 极化SAR图像分类 分类器集成 加权投票准则 

分 类 号：TN957.52[电子电信—信号与信息处理]