基于信息融合的超谱图象分类方法研究

Hyperspectral image classification based on information fusion

针对高数据维给超谱图象有效信息提取和分类造成的困难 ,引入了自适应子空间分解方法进行数据源的划分 ,并在此基础上进行了基于信息融合的超谱图象分类研究 .在根据超谱数据本身特点获得的子空间上进行信息融合 ,有利于分类特征的集中和提取 .实验结果表明 ,利用自适应子空间分解方法划分数据源是有效的 。

High data dimensionality of hyperspectral image results in many difficulties for effective information extraction and classification. Taken into consideration of this, adaptive subspace decomposition (ASD) is introduced in this paper. Then, hyperspectral image classification on the basis of information fusion is studied. Information fusion on the subspaces obtained by ASD is beneficial to the features concentration and extraction. The experiments on AVIRIS images show that it is effective to divide data sources by ASD and feature fusion based on wavelet and decision fusion based on consensus theory are suitable to hyperspectral image classification.