您的位置:首页  > 论文页面

SAR影像特征提取的多纹理方法研究

发表时间:2014-11-30  浏览量:2297  下载量:872
全部作者: 陈军,杜培军,谭琨
作者单位: 中国矿业大学国土环境与灾害监测国家测绘地理信息局重点实验室;南京大学江苏省地理信息技术重点实验室
摘 要: 系统比较了基于主要纹理的合成孔径雷达(synthetic aperture radar,SAR)影像特征提取方法,包括:小波多尺度特征提取方法、地统计学变差函数法、基于分形理论的盒子维提取方法、高斯-马尔可夫特征提取法、灰度共生矩阵提取法、基于概率统计模型的提取方法等。分别应用以上方法及其组合进行SAR影像的纹理特征提取实验,使用分类精度较高的支持向量机(support vector machines, SVM)分类器进行分类。根据分类结果得出结论:1)地对于SAR影像单纹理提取方法,概率统计模型提取法能很好地提取SAR影像的纹理特征;2) 对于两种纹理提取的组合方法——灰度共生矩阵和基于分形理论的盒子维提取方法组合能更好地提取SAR影像的纹理特征。
关 键 词: 摄影测量与遥感技术;特征提取;纹理;概率统计模型;分形理论;灰度共生矩阵
Title: Research of multi-texture extraction methods on SAR image
Author: CHEN Jun, DU Peijun, TAN Kun
Organization: NASG Key Laboratory of Land Environment and Disaster Monitoring, China University of Mining and Technology; Jiangsu Provincial Key Laboratory of Geographic Information Science and Technology, Nanjing University
Abstract: A systemic comparison is done on main texture extraction methods for synthetic aperture radar (SAR) image, including: multi-scale wavelet transform method, variogram of geostatistics method, the fractal methods of box counting, Gaussian Markov mixture model method, gray level co-occurrence matrix method and occurrence model method. These measures as well as their combinations are used to extract textural features from SAR image. Support vector machines (SVM) is used as the classifier for its better performance on classification. Based on these experiments, two conclusions can be drawn: 1) As for single texture extraction methods, occurrence model method is a better way to extract textural features from SAR image; 2) As for two methods’ combination, the union of gray level co-occurrence matrix method and fractal methods of box counting is the best combination to extract textural features from SAR image.
Key words: photogrammetry and remote sensing technology; feature extraction; texture; occurrence model; fractal theory; gray level co-occurrence matrix
发表期数: 2014年11月第22期
引用格式: 陈军,杜培军,谭琨. SAR影像特征提取的多纹理方法研究[J]. 中国科技论文在线精品论文,2014,7(22):2270-2277.
 
0 评论数 0
暂无评论
友情链接