基于层次集成学习的高光谱图像分类

为提高高光谱图像（HSI）分类精度，基于集成学习方法提出高光谱图像分类的层次集成学习新框架。采用两种集成学习策略：外部集成及内部集成。在外部集成阶段，构造多种高光谱图像的光谱和空间特征，使外部集成呈高度多样性，有利于提高分类精度；内部集成阶段，针对关联多特征集中的个体，Adaboost算法实现个体分类性能的提高。两组高光谱数据的实验结果表明，与原始的Adaboost和单分类器相比较，该方法在整体精度方面有更好的性能。

Hierarchy Ensemble Learning for Hyperspectral Imagery Classification

Ensemble learning based methods have demonstrated impressive capacities to improve the classification accuracy of hyperspectral imagery (HSI) .In this paper ,we present a novel hierarchy ensemble learning framework for HSI classification .Hierarchy ensemble is conducted by combing multi‐strategy ensemble in view of ensemble diversity and the accuracy of individual members . In the outer stage ,the allocation of multi‐features to ensemble individuals achieves a high degree of outer ensemble diversity .M ulti‐features of HSI are the integration of diverse spectral and spatial features .In the inner stage ,Adaboost is implemented for each individual in the associated multi‐feature set to improve the individuals’classification performance .Experimental results on two hyperspectral data sets reveal that our proposed method obtains sound performances in terms of better overall accuracies ,compared with original Adaboost and single classifier .