面向对象的高分辨率影像特征选择研究  

Object-oriented features space optimization of high spatial resolution remote sensing image

作　　者：张俊[1] 于庆国[1] 朱晓东[2] 刘恺[1] 

机构地区：[1]国家测绘局第三地形测量队,哈尔滨150081 [2]国家测绘局黑龙江基础地理信息中心,哈尔滨150081

出　　处：《测绘科学》2011年第3期141-143,共3页Science of Surveying and Mapping

基　　金：西部测图项目;国家自然科学基金重点项目(50534050);国家自然科学基金资助项目(50774080)

摘　　要：本文主要研究面向对象的高分辨影像信息提取中的特征选择问题。文中分别选择光谱、纹理、形状等特征57个和28个进行特征优化,得到两组分别由46个和4个特征组成的不同的最优特征集,并利用这些特征集采用K近邻、模糊与K近邻级联两种不同的面向对象分类策略进行分类研究。最后从合理性、效率和精度三方面进行了对比分析。实验结果表明,对所有类别进行特征空间优化费时费力,并不合适,分类时需要采用级联的方式将两种分类器联合使用,根据实际需要进行必要的特征选择。In this paper,object-oriented features space optimization of high spatial resolution remote sensing image classification was discussed.Two different groups of features space were optimized by 46 and 4 features from 57 and 28 disparate texture,shape and structure information of image objects,and were used to extract information by two different object-oriented classification strategies： K nearest neighbor classification,cascade connection of fuzzy classification and K nearest neighbor classification.The experiment results were compared on the aspects of rationality,efficiency and precision,which showed that features space optimization of each classification was time-consuming and improper,and the extraction effect of cascade connection classification strategy increased significantly.

关 键 词：高分辨率遥感 面向对象 K近邻 特征优化 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]