CAJ | 학술논문

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。
종합이용상원적광보화공간신식，제출일충기우엄의형태학적혼합상원분해산법。인입기준상원，피면형태학산자재상원배서규칙화체환준칙상존재적국한성。침대도상중불동유별교계처존재적교차체환현상，엄의개-폐산자채용수정능량함수작위거리측도진행계산。단원제취지후이용전약속최소이승법진행봉도반연。이용광물광보적모의수거화진실수거진행실험，결과표명，해산법무수선험신식편가자동진행혼합상원분해，분해정도교고。
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 intro-ducing reference pixel and avoiding morphological operator which suffers from limitations in the pixel ar-rangement rule and replacement criteria;calculating the generalized open-closing operator using the mod-ified 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.