摘要
针对高光谱数据维数高、数据量大、信息冗余多、波段相关性强等特点,在综合各种数据降维方法的基础上,提出一种基于最佳波段组合的高光谱遥感影像分类方法。以美国印第安纳州地区的AVIRIS数据为例,分析各波段信息量和相邻波段的相关性,利用子空间划分、分段波段指数选择法,进行特征波段的选择;并针对难区分地物类别,应用J-M距离模型对其可分性进行判别,获得最佳波段组合。最后采用支持向量机分类器进行分类。实验结果表明,采用最佳波段组合方法,可以有效地提高高光谱的分类精度。
Comparing with traditional remote sensing data, high dimension , the large amount of information , data redundancy and strong band relevant are prevalent in the hyperspetral remote sensing data .On the combination of data dimension reduction methods , this paper puts forward a hyperspectral remote sensing image classification method based on optimum band combination .Taking AVIRIS data in Indiana area as an example , we analyze the spectrum information and the adjacent band correlation .Then we use sub-space division and band index , select the feature band .For the classes difficult to distinguish , we apply J-M distance model for its separability criterion to point out the optimal band combination .Finally, we use SVM classifier for classification .The experimental re-sults show that using the optimal band combination can effectively improve the classification accuracy of hyperspectral remote sensing data.
出处
《测绘与空间地理信息》
2014年第4期19-22,共4页
Geomatics & Spatial Information Technology
