摘要
为解决传统合成孔径雷达(SAR)图像目视解译的困难,对一种基于信息可视化技术的SAR图像目视解译方法进行了研究。基于非线性流形学习理论的维数约减技术,对原始单极化和多极化SAR图像进行特征提取和数据挖掘,通过基于极化数据变换的特征和基于极化目标分解获取SAR图像本征特征,选择利于用户应用的特征在彩色空间编码重构出SAR信息图像提供给判读员进行解译。给出了基于信息可视化技术的SAR图像目视解译框架。多种应用结果表明:该法能挖掘并显示出大量图像中隐含的信息,产生的特征图像较原始SAR图像更符合人眼视觉,可有效解译SAR图像。
To solve the difficulty problem in traditional synthetic aperture radar(SAR)image visual interpretation,a visual interpretation method for SAR images based on information visualization was studied in this paper.The feature extraction and data mining had been carried out on original SAR images using dimensionality reduction technology based on nonlinear manifold learning theory.The SAR image intrinsic features which were beneficial for application were obtained through the features based on polarimetric data transfer and polarimetric target decomposition.The intrinsic features could be normalized in color space,and visualized to interpreters for visual interpretation.The framework of SAR image visual interpretation based on information visualization was given out.The various application results showed that this method could utilize large amount of information hidden in original SAR images.The intrinsic feature images were better for human vision and effective to visual interpretation for SAR images.
出处
《上海航天》
2016年第4期81-87,共7页
Aerospace Shanghai
基金
国家自然科学基金资助(41501414)
关键词
合成孔径雷达
SAR图像
目视解译
特征提取
信息可视化
流形学习
维数约减
拉普拉斯特征映射
Synthetic aperture radar
SAR image
Visual interpretation
Feature extraction
Information visualization
Manifold learning
Dimensionnality reduction
Laplacian eigenmaps
