| 归一化地物子空间投影下的光谱解混方法 |
| |
| 引用本文： | 许承权,邓雪彬.归一化地物子空间投影下的光谱解混方法[J].测绘科学,2021,46(3):117-123. |
| |
| 作者姓名： | 许承权 邓雪彬 |
| |
| 作者单位： | 闽江学院海洋学院,福州350108;国家知识产权局专利局专利审查协作湖北中心,武汉430075 |
| |
| 基金项目： | 国家自然科学基金项目(41404008);福建省自然科学基金项目(2020J01834)。 |
| |
| 摘 要： | 针对线性光谱解混方法,全约束条件下的最小二乘准则和正交子空间投影(OSP),因缺乏物理约束条件使得组分丰度估值容易出现负值这一问题,该文在线性光谱混合分析模型中增加光谱组分丰度"和为1"且为"非负"的约束条件,提出了归一化地物子空间投影下(NMSP)的光谱解混方法。该方法假定一条基准端元已知以消除组分之间的相关性,再基于基准端元对端元矩阵和影像矩阵进行平移,进一步消除像元在端元方向投影时原点引起的错误。实验结果表明,与约束条件下的OSP分类器以及最小二乘法相比,NMSP在光谱解混中可以得到更加合理的地物组分丰度且能保持端元丰度"非负"和稀疏的物理特性。
|
| 关 键 词： | 归一化地物子空间投影 正交子空间投影 最小二乘法 线性光谱混合分析 |
Methods for spectral unmixing under normalized materials subspace projection |
| |
| XU Chengquan,DENG Xuebin.Methods for spectral unmixing under normalized materials subspace projection[J].Science of Surveying and Mapping,2021,46(3):117-123. |
| |
| Authors: | XU Chengquan DENG Xuebin |
| |
| Institution: | (Ocean College,Minjiang University,Fuzhou 350108,China;Patent Examination Cooperation Hubei Center of the Patent Office,CNIPA,Wuhan 430075,China) |
| |
| Abstract: | Aiming at the problem that the main methods of linear spectral unmixing,the least square criterion and orthogonal subspace projection(OSP)under full constraint conditions,the endmember abundance estimation was prone to negative value due to the lack of physical constraints,the constraint conditions that the spectral endmember abundances"sum to 1"and"non negative"were added to the linear spectral mixture analysis model,and a Normalized Materials Subspace Projection(NMSP)was proposed in this paper.The correlation among endmembers was eliminated with the assist of one known endmember.Both endmembers and image matrix were translated by the known base endmember so that the error caused by the origin of coordinates were excluded when image pixels were projected in endmembers directions.The results showed that NMSP was able to achieve a more reasonable and reliable material fraction abundance in spectral unmixing compared to fully constrained OSP and least squares method.Nonnegative and sparse physical characteristics were well kept in endmember fraction by NMSP. |
| |
| Keywords: | NMSP OSP least squares method linear spectral unmixing analysis |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
|
