基于广义形态学的高光谱图像解混算法  

Generalized morphology-based hyperspectral image unmixing algorithm

作　　者：赵岩[1] 张春晶[2] 王东辉[3] 

机构地区：[1]黑龙江科技大学电气与控制工程学院,哈尔滨150022 [2]黑龙江科技大学电子与信息工程学院,哈尔滨150022 [3]哈尔滨工程大学信息与通信工程学院,哈尔滨150001

出　　处：《黑龙江科技大学学报》2015年第3期343-347,共5页Journal of Heilongjiang University of Science And Technology

基　　金：黑龙江省教育厅科学技术研究面上项目(12541734)

摘　　要：综合利用像元的光谱和空间信息,提出一种基于广义形态学的混合像元分解算法。引入基准像元,避免形态学算子在像元排序规则和替换准则上存在的局限性。针对图像中不同类别交界处存在的交叉替换现象,广义开-闭算子采用修正能量函数作为距离测度进行计算。端元提取之后利用全约束最小二乘法进行丰度反演。利用矿物光谱的模拟数据和真实数据进行实验,结果表明,该算法无需先验信息便可自动进行混合像元分解,分解精度较高。This paper introduces an unmixing algorithm based on generalized morphology,thanks to comprehensive utilization of pixel spectral and spatial information. The algorithm study consists of introducing reference pixel and avoiding morphological operator which suffers from limitations in the pixel arrangement rule and replacement criteria; calculating the generalized open-closing operator using the modified energy function as a distance measure in response to cross substitution occurring in different types of junction in the image; and performing abundance inversion using fully constrainded least squares,after the endmember extraction. The experiment using simulated data and real data of mineral spectrum shows that proposed algorithm has a demonstrated ability to automatically decompose mixed pixel without prior information and exhibits a higher decomposition precision.

关 键 词：高光谱混合像元分解 广义形态学 修正能量函数 

分 类 号：TN911.73[电子电信—通信与信息系统]