面向应用的高光谱影像分类方法

An Application-Oriented Hyperspectral Classification Scheme

提出了一种面向应用的高光谱影像分类方法,旨在从根本上、全方位地削弱各种不利因素对该类影像分类精度的影响。主要包括利用IEM算法获取更为精确的类别分布信息,采用Tabu搜索算法进行原始特征空间的降维,运用基于混合规则的组合分类器来判断待识样本的类别标签。实验表明,按照该方法进行高光谱影像的分类处理,可以得到很高精度的分类结果。

In this paper, an application-oriented hyperspectral classification scheme is proposed. The scheme includes three key parts： the utilization of IEM algorithm to obtain more accurate classes estimation, the utilization of Tabu search algorithm to reduce the number of the input dimensionality, the utilization of combined classifiers based on hybrid combing rule to assign an appropriate class label to the input sample. The experiment demonstrates that the proposed hyperspectral classification is very effective and robust.