一种基于局部判别正切空间排列的高光谱遥感影像降维方法

摘要

A local discriminative tangent space alignment （LDTSA） based dimension reduction method is proposed, It applies the idea of part optimization and whole alignment and considers encoding the geometric and discrimina- tive information in a local patch, The experiment demonstrate the effectiveness of LDTSA compared with represent- ative dimensionality reduction algorithms, O LDTSA avoids the small-sample-size problern② LDTSA preserves the discriminative ability; ③LDTSA has the ability to detect the intrinsic structure from the hyperspectral data.%提出一种基于局部判别正切空间排列（iocal discriminative tangent space alignment，LDTSA）的高光谱影像降维方法。LDTSA源于局部正切空间排列（LTSA）中的排列机制，在一个局域块内利用线性局部正切平面对类内样本的流形结构建模，同时还考虑到类间判别信息以最大化判别边界。利用多幅高光谱数据进行降维和分类试验。结果表明，LDTSA主要有三个优点：①在小样本问题上性能稳定；②在降维过程中保持类别间的判别信息；③有效挖掘数据集的几何流形结构。