激光与光电子学进展, 2019, 56 (9): 091001, 网络出版: 2019-07-05
基于数据简化的改进非负矩阵分解端元提取方法 下载: 943次
Improved Algorithm for Nonnegative Matrix Factorization and Endmember Extraction Based on Data Simplification
图像处理 高光谱遥感 混合像元分解 光谱信息熵 非负矩阵分解 端元提取 image processing hyperspectral remote sensing mixed pixel decomposition spectral information entropy nonnegative matrix decomposition endmember extraction
摘要
提出了一种基于高光谱数据简化的改进非负矩阵分解端元提取方法,通过计算和比较图像的光谱信息熵,划分图像的同质区,只选择同质区中最具代表性的像元参与非负矩阵分解运算,减少了端元提取算法的运算量。实验结果显示,数据简化前后运用非负矩阵分解算法所提取的几种矿物的光谱角均值基本相等,但数据简化后端元提取算法的运行时间减少了4/5,算法的运行效率提高。
Abstract
An improved method for nonnegative matrix decomposition and endmember extraction is proposed based on hyperspectral data simplification. Further, the homogeneous regions of images can be identified by calculating and comparing the spectral information entropy of various regions. Only the most representative pixels in the homogeneous regions are selected for application in the subsequent nonnegative matrix decomposition algorithm, which considerably reduces the amount of computation required in the endmember extraction algorithm. The experimental results show that although the mean values of the spectral angles of several kinds of minerals extracted using the nonnegative matrix factorization algorithm before and after data simplification are equal, the operation time of endmember extraction after data simplification is reduced by approximately 4/5, and the operating efficiency of the algorithm is improved.
徐君, 王旭红, 王彩玲. 基于数据简化的改进非负矩阵分解端元提取方法[J]. 激光与光电子学进展, 2019, 56(9): 091001. Jun Xu, Xuhong Wang, Cailing Wang. Improved Algorithm for Nonnegative Matrix Factorization and Endmember Extraction Based on Data Simplification[J]. Laser & Optoelectronics Progress, 2019, 56(9): 091001.