散射项约束非负矩阵分解的高光谱图像解混  

Scattering-term constrained nonnegative matrix factorization for hyperspectral unmixing

作　　者：陈善学[1,2] 董桓宇 陈雯雯 CHEN Shanxue;DONG Huanyu;CHEN Wenwen(Chongqing University of Posts and Telecommunications,School of Communication and Information Engineering,Chongqing 400065,P.R.China;Chongqing Key Lab of Mobile Communications Technology,Chongqing University of Posts and Telecommunications,Chongqing 400065,P.R.China)

机构地区：[1]重庆邮电大学通信与信息工程学院,重庆400065 [2]重庆邮电大学移动通信技术重庆市重点实验室,重庆400065

出　　处：《重庆邮电大学学报（自然科学版）》2021年第2期296-303,共8页Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)

基　　金：国家自然科学基金(61771084,61102062);重庆市科委自然科学基金(cstc2015jcyjA40050)。

摘　　要：针对传统的约束非负矩阵分解方法对于解混的物理特性考虑较少,提出一种高光谱图像的解混方法:散射项约束非负矩阵分解(scattering-term constrained nonnegative matrix factorization, STC-NMF)。与大多数约束非负矩阵分解算法将约束建立在数据的数学特性之上不同,STC-NMF考虑到大气中的悬浮物、胶着物的散射作用,对成像光谱仪接收到的光谱信号有着不可忽视的影响,将大气中的米氏散射(Mie scattering)造成的邻域贡献视作干扰,通过将散射相函数作为约束条件的约束非负矩阵分解,在丰度上对目标像素及其邻域的米氏散射干扰进行约束,以达到在目标函数上将米氏散射和噪声造成的干扰降低到最小的效果,通过实验验证了提出方法的有效性。In view of the fact that the traditional constrained nonnegative matrix factorization seldom considers the physical characteristics of the unmixed objects,this paper proposes a method for hyperspectral unmixing:scattering-term constrained nonnegative matrix factorization(STC-NMF).Different from the most constrained nonnegative matrix factorization,STC-NMF considers the scattering of suspended matter and adhesive matter in atmosphere which has a nonnegligible impact on the spectral signals received by the imaging spectrometer.The neighborhood contribution caused by Mie scattering in the atmosphere is regarded as interference,so the Scattering-phase Function is regarded as constraint non-negative matrix factorization.And this constraint constrains the interference of Mie Scattering of the target pixel and its neighborhood in terms of abundance,in order to minimize the interference caused by Mie scattering and noise in the objective function,and the effectiveness of the proposed method is verified by experiments.

关 键 词：高光谱图像解混 非负矩阵分解(NMF) 散射相函数 米氏散射 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]