摘要
研究遥感图像处理混合像元分解问题,由于遥感图像中包含很多混合像元,影响图像的质量。传统的解混算法是以线性光谱混合模型为基础,因此解混精度不高。针对遥感图像中存在的非线性光谱混合,从数据统计的角度,提出一种基于光谱夹角距离的局部切空间排列算法进行非线性降维的光谱解混算法。首先通过基于光谱夹角距离的局部切空间排列算法将原始高光谱数据非线性降维到低维空间,再利用寻找最大单形体体积的方法提取端元,并用非负约束的最小二乘法计算各个端元的丰度。通过仿真和真实高光谱遥感数据实验结果表明,得到的分解结果优于测地线的最大单形体体积(GSVM)算法和N-FINDR算法,为提高遥感图像分解精度提供了参考。
The existence of mixed pixels influences the qualities of remote sensing images, so spectral unmixing was researched in this paper. Since traditional spectral unmixing algorithms are based on linear spectral mixture mod el, they perform poorly in unmixing mixed pixel. Because mixed pixel in hyperspectral image is actually nonlinear mixing of endmembers, a new data driven method was proposed by using of local tangent space alignment based on spectral angel distance in this paper. The proposed method was adapted to reduce original data into a low dimen sional space. End members were extracted by looking for the largest simplex volume from low dimensional space. The abundances of endmembers were calculated by non negative constrained least squares. The experimental results of synthetic mixtures and a real image scene demonstrat that the accuracy of proposed method outperformes the geo desic simplex volume maximization (GSVM) and N FINDR algorithm, and the calculation speed is faster than GS VM and close to N FINDR algorithm, which provides some reference for improving the accuracy of remote sensing image analysis .
出处
《计算机仿真》
CSCD
北大核心
2014年第4期347-351,共5页
Computer Simulation
基金
国家自然科学基金资助项目(60702017)
航空科学基金(20100112002)
关键词
混合像元
高光谱解混
非线性降维
局部切空间排列
流形学习
Mixed pixel
Hyperspectral unmixing
Nonlinear dimensionality reduction
Local tangent space align-ment
Manifold learning