High-Level Feature Selection With Dictionary Learning for Unsupervised SAR Imagery Terrain Classification
High-Level Feature Selection With Dictionary Learning for Unsupervised SAR Imagery Terrain Classification
复制标题
用于无监督 SAR 图像地形分类的字典学习高级特征选择
DOI：
10.1109/jstars.2016.2530850
复制
发表时间：
2017
影响因子：
5.5
通讯作者：
Wen Zaidao
中科院分区：
文献类型：
--
作者：
Chen Jiawei;Jiao Licheng;Wen Zaidao
合成孔径雷达（SAR）图像地物分类是一项重要的研究内容，但低层特征容易受到斑点噪声的影响，无法或不能准确地捕捉到复杂的、不规则的纹理结构。在本文中，提出了一种新的特征学习框架来解决这个问题，其中一些中级和高级功能，同时学习，利用空间上下文约束和稀疏先验。更具体地说，作为中间层的中间层特征是通过空间约束从初始化的几个低层特征中提取出来的，以减少斑点噪声的影响。然后，通过有效的字典学习算法学习更抽象、更有鉴别力的高层特征，以表示SAR图像中的复杂结构。最后，人工合成和真实的SAR图像被用来验证所提出的框架的有效性。实验结果表明，该算法的分类性能优于其他同类算法，且所学习的高级特征对斑点噪声具有较强的鲁棒性，能够提高分类性能。
Features are of great importance for synthetic aperture radar (SAR) imagery terrain classification, but low-level features usually readily suffer from the speckle noise and they are incapable or inaccurate to capture some complex and irregular texture structure. In this paper, a novel feature learning framework is proposed to address this problem, in which some mid-level and high-level features are simultaneously learned by exploiting the spatial context constraints and sparse priors. More specifically, the mid-level features served as the intermediates are extracted from several initialized low-level features by the spatial constraints to reduce the influence of the speckle noise. Then, more abstract and discriminative high-level features are learned with an effective dictionary learning algorithm so as to represent the complex structures in SAR imagery. Finally, both artificial synthesis and real SAR imagery are utilized to verify the effectiveness of the proposed framework. It is demonstrated from both quantitative evaluations and visual results that the proposed algorithm performs better than other compared algorithms and the learned high-level feature is robust to the speckle noise and can improve the classification performance.