高光谱遥感图像光谱加权稀疏解混算法分析和比较  

Analysis and comparison of spectral weighted sparse unmixing algorithms of hyperspectral remote sensing images

作　　者：徐晨光[1] 张绍泉 李璠[1] 吴朝明 XU Chenguang;ZHANG Shaoquan;LI Fan;WU Zhaoming(Jiangxi Province Key Laboratory of Water Information Cooperative Sensing and Intelligent Processing,Nanchang 330099,China)

机构地区：[1]南昌工程学院江西省水信息协同感知与智能处理重点实验室,江西南昌330099

出　　处：《南昌工程学院学报》2020年第6期87-93,共7页Journal of Nanchang Institute of Technology

基　　金：江西省教育厅科学技术研究项目(GJJ180963);江西省科技厅重点研发项目(20202BBGL73081)。

摘　　要：由于受到高光谱传感器空间分辨率的限制以及复杂地物的影响,遥感影像中的一个像元所包含的端元信息并不单一,普遍存在着混合像元的情况。如何把混合像元分解出来,对高光谱影像进行解混,直接影响着后续地物目标的识别与分类。高光谱稀疏解混算法是在庞大的光谱库中进行光谱端元的筛选并对其丰度进行估计的高光谱解混技术。近年来高光谱稀疏解混方法受到广泛的关注,并且取得了一定的成果。本文主要介绍了几种经典的稀疏解混算法和光谱加权的稀疏解混算法的原理和优缺点,并把几种算法进行了详细的分析和比较。通过模拟和真实数据实验验证了几种算法的解混效果。Due to the limitation of spatial resolution of hyperspectral sensors and the influence of complex ground features,the endmember information contained in one pixel of remote sensing image is not single,and the mixed pixels are common.The ways of decomposing the mixed pixels and unmixing the hyperspectral images directly affect the recognition and classification of the following ground objects.The hyperspectral thinning and unmixing algorithm is a hyperspectral unmixing technique used to screen spectral endmembers and estimate their abundance in a huge spectral library.In recent years,the hyperspectral sparse unmixing method has been widely concerned,and some achievements have been made.This paper mainly introduces the principles,advantages and disadvantages of several classical sparse unmixing algorithms and spectral weighted sparse unmixing algorithms,and analyzes and compares them in detail.The results of simulation and real data experiments verify the unmixing effect of the algorithms.

关 键 词：高光谱图像 混合像元 稀疏解混 光谱加权 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]