摘要
以支持向量数和相关性分析为评估依据,结合序列前进搜寻策略,本文提出一种顾及特征优化的改进SVM分类方法,并将其应用于全极化SAR图像监督分类。真实数据的实验结果表明,该方法不仅具有小样本情况下的良好泛化性能,而且能以更少的特征个数,在更广泛的SVM参数取值范围内获得更高的分类精度。
An improved SVM classification method for fully polarimetric SAR imagery concerned with feature optimization was pro- posed in this paper, which is based on the assessment of support vector number and correlation analysis, and combining with SFS search strategy. Experimental results showed that the proposed method could not only retain good generalization with limited samples, but also obtain higher classification accuracy with less number of features in a wider range of SVM parameters.
出处
《测绘科学》
CSCD
北大核心
2013年第3期115-117,139,共4页
Science of Surveying and Mapping
基金
国家863计划资助项目(2007AA12Z143)
国家自然科学基金资助项目(40201039
40771157
41001260)
关键词
极化SAR
特征优化
监督分类
支持向量机
polarimetric Synthetic Aperture Radar
feature optimization
supervised classification
Support Vector Machine
