基于Tabu搜索的高光谱影像特征选择  

Feature Selection in Hyperspectral Classification Based on Tabu Search Algorithm

作　　者：杨哲海[1] 张雅争[2] 宫大鹏 李之歆 韩建峰 

机构地区：[1]解放军信息工程大学测绘学院 [2]海军大连舰艇学院 [3]65015部队

出　　处：《海洋测绘》2006年第4期11-14,共4页Hydrographic Surveying and Charting

基　　金：国家863计划资助项目(2001AA131090)

摘　　要：在高光谱影像的分类过程中,如何有效地降低特征空间的维数,又能保证原始数据所包含的丰富地物信息,是一项十分重要而繁琐的工作。在分析了传统降维方法所面临问题的基础上,将禁忌(Tabu)搜索算法引入到高光谱影像的特征选择研究,指出由于Tabu搜索算法所具有的良好全局寻优能力,因而在该类影像的降维研究中有着广阔的应用前景。实验表明,将Tabu搜索算法获取的波段进行高光谱影像分类,在求解的时间和分类结果的精度上都可达到令人满意的效果。In hyperspectral classification,it is necessary to reduce both the number of input dimensionality and maintain the information contained in original hyperspectral data. Based on analyzing the limitations of the traditional dimensionality reduction methods, the paper introduced the Tabu Search into hyperspectral feature selection. Also, the author pointed out that due to the advantage of solving the global optimal problems, Tabu Search algorithm will be widely used in hyperspectral feature reduction. Experimental results demonstrate that Tabu Search is very helpful to achieve good classification performance by selecting the most useful bands which will be used in classifiers.

关 键 词：高光谱 遥感 特征选择 禁忌搜索 

分 类 号：P237[天文地球—摄影测量与遥感]