结合视觉特征的极化SAR图像分类  

Integrating visual features in polarimetric SAR image classification

作　　者：黄鹏艳 卜丽静[2] 范永良 HUANG Pengyan;BU Lijing;FAN Yongliang(School of Civil Engineering, Luoyang Institute of Science and Technology, Luoyang 471000, China;School of Mapping and Geographical Science, Liaoning Technical University, Fuxin 123000,China;College of Civil Engineering and Mechanics, Lanzhou University, Lanzhou 730000, China)

机构地区：[1]洛阳理工学院土木工程学院,洛阳471000 [2]辽宁工程技术大学测绘与地理科学学院,阜新123000 [3]兰州大学土木工程与力学学院,兰州730000

出　　处：《国土资源遥感》2020年第2期88-93,共6页Remote Sensing for Land & Resources

基　　金：国家自然科学基金青年科学基金项目“增量式非参局部自适应MAP模型的凝视卫星视频运动场景超分辨率重建”(编号41801294);洛阳理工学院青年基金研究项目“极化SAR图像的分类方法研究”(编号2016QZ06)共同资助。

摘　　要：为充分提取极化合成孔径雷达(synthetic aperture Radar,SAR)图像中的信息,提高图像分类精度,提出结合视觉特征的极化SAR图像分类方法。首先,通过极化目标分解方法提取极化参数组成极化特征向量;然后,通过灰度共生矩阵和假彩色合成图像提取极化SAR图像中的纹理和颜色特征参数构成视觉特征向量;最后,将视觉特征向量与极化特征向量组合成新的特征向量,并利用支持向量机(support vector machine,SVM)方法进行分类。对RADARSAT-2的全极化SLC数据进行分类实验,结果表明,与仅使用极化特征向量相比,视觉特征的加入能有效提高极化SAR图像的分类精度。In order to improve the polarimetric synthetic aperture Radar(SAR)images classification accuracy by fully extracting variety of useful information,this paper proposes integrating visual features in SAR images classification.Firstly the authors constructed the polarimetric decomposition feature vector,then extracted texture parameters with Grayscale symbiosis matrix,and finally extracted color feature parameters by pseudo-color image.Based on constructing visual vector with texture and color parameter,the authors integrated the visual vector with the polarimetric feature vector to combine the new feature vectors.Using different feature vectors for classification of full PolSAR image,the authors made a comparative study of the classification results.The results show that the combination of visual features can effectively improve the classification accuracy of fully polarimetric SAR image.

关 键 词：目标分解 视觉特征 极化SAR分类 特征向量 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]