基于纹理特征的高分辨率影像城郊居民区提取方法研究  

The Research on Suburb Residential Areas Extraction of High Spatial Resolution Remotely Sensed Imagery Based on Texture Features

作　　者：许锐[1] 

机构地区：[1]福建工程学院信息科学与工程学院,福州350108

出　　处：《科学技术与工程》2013年第33期10017-10020,共4页Science Technology and Engineering

基　　金：国家自然科学基金资助项目(40871206);福建省教育厅科技研究资助项目(JA13220;JB11116)资助

摘　　要：高空间分辨率遥感数据由于存在大量同物异谱和异物同谱现象。传统的方法难以对城郊的居民区取得满意的提取效果。在高空间分辨率遥感影像对象提取研究中引入子图表示方法,通过综合Contourlet变换的多尺度、多方向分解能力,以及灰度共生矩阵的统计能力,实现纹理特征的合理表达,较好地描述遥感影像对象的纹理特征。利用多核学习,得到反应各特征区分能力的特征权值。实验以QuickBird影像为数据样本,实验结果表明提出算法有效地实现城郊居民区提取。Traditional extraction algorithmof residential areaextraction in suburenvironmendo nogive the desired resuldue to large within-classpectral variationand between-classpectral confusionthacharacterize the high spatial resolution remotely sensed data. The Subgraph Method waintroduced foobjectextraction from high spatial resolution remotely sensed imagery. Reasonable expression of texture featurewaachieved by integrating the multi-scale multi-directional capabilitieof Contourletransform and the statistical capacity of GLCM, which can correctly describe the featureof remote sensing imageobjects. Weightto show distinguishing ability of each fea- ture wacalculated by the Multiple kernel learning（MKL） methods. case study taking QucikBird imagery asample data, provethe effectivenesof the innovative method adopted in thiresearch.

关 键 词：遥感影像 纹理特征 CONTOURLET变换 灰度共生矩阵 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]