基于随机子空间核极端学习机集成的高光谱遥感图像分类

结合随机子空间和核极端学习机集成提出了一种新的高光谱遥感图像分类方法。首先利用随机子空间方法从高光谱遥感图像数据的整体特征中随机生成多个大小相同的特征子集；然后利用核极端学习机在这些特征子集上进行训练从而获得基分类器；最后将所有基分类器的输出集成起来,通过投票机制得到分类结果。在高光谱遥感图像数据集上的实验结果表明:所提方法能够提高分类效果,且其分类总精度要高于核极端学习机和随机森林方法。

Classification of Hyperspectral Remote Sensing Image Based on Random Subspace and Kernel Extreme Learning Machine Ensemble

This paper presented a novel classification algorithm of hyperspectral remote sensing image based on random subspace and kernel extreme learning machine ensemble.Firstly,many feature subsets of the same size are generated from the whole feature of hyperspectral remote sensing image data with random subspace method.Then the base classifiers of kernel extreme learning machine are trained based on these feature subsets.Finally,the classification result is decided by voting strategy.The experimental results on hyperspectral remote sensing image indicate that the proposed method has better performance than the methods based on kernel extreme learning machine and random forest respectively,and has a higher classification overall accuracy.