等距映射降维用于高光谱影像低维流形特征提取
Low-dimension Manifold Feature Extraction of Hyperspectral Imagery Using Dimension Reduction with Isomap
-
摘要: 通过观察对比等距映射流形坐标和光谱特征的变化趋势来解释每一维流形坐标的光谱含义,目的在于从具有光谱解释的流形图中提取低维流形特征。通过设计两个实例来验证本文提出的低维流形特征提取方法。结果显示,Isomap低维流形图可用于提取目标地物的低维流形特征,同时也证明了等距映射流形坐标光谱解释的可行性。这对Isomap降维在高光谱影像中的应用具有很大的理论指导意义。Abstract: Nonlinear dimensionality reduction on hyperspectral imagery can be achieved using the isometric mapping(Isomap) method.We explore the spectral interpretations of Isomap manifold coordinates through observing and comparing the changing trends between manifold coordinates and spectral signatures.The study aims to extract desired low-dimension manifold features from Isomap manifold maps.Two cases study are designed to testify the capacity of manifold maps in extracting low-dimension manifold features.The results show that the Isomap manifold maps can be used to extract low-dimension manifold features.Moreover,the results prove that the spectral interpretations of manifold coordinates are feasible.This will be helpful for the applications of the Isomap method in hyperspectral imagery fields.
下载:
