贝叶斯集成框架下的极化SAR图像分类

发布时间：2018-06-20 00:28

本文选题：极化合成孔径雷达 + 图像分类　；参考：《西安电子科技大学学报》2015年02期

【摘要】：对于极化合成孔径雷达(SAR)图像,由于雷达角度和地物形状导致属于同一类的数据可能存在较大的差异性.针对此问题提出了一种基于贝叶斯集成框架的极化SAR图像分类方法.该算法采用贝叶斯集成,通过学习不同个体而获得的分类面来改善极化SAR图像分类性能.首先,输入极化SAR图像,并获得其对应的极化SAR数据及特征.从图像的每一类中任意选择像素点作为图像分类的原始训练样本,并对其进行随机划分获得不同的样本子集.然后,基于获得的样本子集构造对应极化SAR图像的贝叶斯集成框架.最后,通过构造的贝叶斯集成框架对极化SAR图像进行分类.特别在构造贝叶斯集成框架中采用支撑矢量机作为基本的分类器算法.实验结果表明,所提出的算法相比经典的极化SAR分类方法和单个SVM的极化SAR分类方法获得更好的分类性能.
[Abstract]:For polarimetric synthetic Aperture Radar (SAR) images, there may be large differences in data belonging to the same class due to radar angle and object shape. To solve this problem, a polarimetric SAR image classification method based on Bayesian integrated framework is proposed. Bayesian ensemble is used to improve the classification performance of polarimetric SAR images by learning the classification surfaces obtained from different individuals. Firstly, the polarimetric SAR image is input and the corresponding polarimetric SAR data and features are obtained. The pixel points are chosen as the original training samples of image classification from each class of images, and different subsets of samples are obtained by randomly partitioning them. Then, a Bayesian ensemble framework for polarimetric SAR images is constructed based on the obtained sample subsets. Finally, the polarimetric SAR images are classified by a Bayesian integrated framework. In particular, support vector machine (SVM) is used as the basic classifier algorithm in constructing Bayesian integration framework. Experimental results show that the proposed algorithm has better classification performance than classical polarimetric SAR classification method and single SVM polarimetric SAR classification method.
【作者单位】：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;西安电子科技大学计算机学院;
【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) 国家教育部博士点基金资助项目(20100203120005)
【分类号】：TN957.52

【参考文献】