互联网 qkzz.net
全刊杂志网:首页 > 女性 > 文章正文
刊社推荐

基于最速上升算法的超光谱图像波段选择搜索算法周 爽 张钧萍 苏宝库


   (哈尔滨工业大学 a.电子与信息技术研究院;b.航天学院, 哈尔滨 150001)

  

  摘要:超光谱遥感数据具有的波段数目多、波段宽度窄、数据量庞大等特点,给图像的进一步解译带来困难。结合超光谱图像波段选择的具体应用,根据波段之间的相关性将整个波段划分为几个子波段,采用最速上升的特征选择搜索算法在各子波段中快速提取最优波段。为了验证本算法的有效性,分别选取jm距离、bh距离以及类内类间离散度作为评价准则,针对一幅200波段的aviris超光谱图像进行分类实验,并将该方法与传统的sffs算法进行对比。实验结果表明所采用的算法用于特征选择具有搜索能力强、分类精度高的特点,完全可以替代传统的sffs算法。

  关键词:超光谱图像;特征选择;最速上升搜索算法

  中图分类号:tp391文献标志码:a

  文章编号:1001-3695(2008)11-3501-03

  

  band selection algorithm of hyperspectral imagebased on steepest-ascent search algorithm

  

  zhou shuanga,b,zhang jun-pinga,su bao-kub

  (a.school of electronics & information technology,b.school of astronautics, harbininstitute of technology, harbin 150001, china)abstract:this paper proposed a band selection approach of hyperspectral image based on steepest-ascent search algorithm. the approach needed to divide the whole hyperspectral band into several subgroups in terms of the relativity between bands firstly, and then applied the steepest-ascent search strategy to quickly extracting optimal band in every subgroup in which the combinations of bands was indicated by binary vectors and the search was being along the steepest direction until the local extreme was acquired. in order to verify the ralidity of this algorithm, the approach was compared with the classical sequential forward floating selection suboptimal techniques, using hyperspectral remote sensing images as a data set. experimental results prove the ralidity of this algorithms, which can be regarded as a valid alternative to classical sffs method.

......
很抱歉,暂无全文,若需要阅读全文或喜欢本刊物请联系《计算机应用研究》杂志社购买。
欢迎作者提供全文,请点击编辑
分享:
 

了解更多资讯,请关注“木兰百花园”
分享:
 
精彩图文


关键字
支持中国杂志产业发展,请购买、订阅纸质杂志,欢迎杂志社提供过刊、样刊及电子版。
关于我们 | 网站声明 | 刊社管理 | 网站地图 | 联系方式 | 中图分类法 | RSS 2.0订阅 | IP查询
全刊杂志赏析网 2017