高光谱影像分类及组合分类器的应用  

Hyperspectral Classification and the Utilization of Combined Classifiers

作　　者：宫大鹏 李之歆 韩建峰 杨哲海[1] 张雅争[2] 

机构地区：[1]65015部队 [2]海军大连舰艇学院

出　　处：《海洋测绘》2006年第5期20-23,共4页Hydrographic Surveying and Charting

基　　金：国家863计划(2001AA135160)

摘　　要：高光谱影像包含有丰富的地物类别信息,但由于受分类方法等因素的限制,这些信息并不能得到充分的挖掘。组合分类器由于可以实现各分类结果信息的互补,因而总有着较各单分类器效果更好的分类性能。将组合分类方法引入到高光谱影像的分类实践中,并对组合的关键问题一组合规则进行探讨,提出了混合组合规则的概念。实验表明,利用该组合分类方法对高光谱影像进行分类,可以得到良好的处理结果。Hyperspectral images can provide more information about the earth＇s surface due to the higher spectral resolution. Unfortunately the classification performance will be deteriorated with the liniations of the classifiers. Nevertheless, combined classifiers, which utilize the comlementary information contained in the classification results, usually leads to much better classification performance. In this paper the combined classifiers were introduced into the hyperspectral classification. Also a new hybrid combination rule was proposed. Experiment demonstrated that the proposed combined classifiers are superior in hyperspectral classification.

关 键 词：高光谱 分类 分类器 组合 

分 类 号：P237[天文地球—摄影测量与遥感]