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

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 .