多源遥感图像融合的数据同化算法  

Data Assimilation Algorithm of Multi-fountain Remote Sensing Image Fusion

作　　者：付炜[1] 裴欢[1] 廖晓玉[1] 白超[1] 高现文[1] 田贺[1] 朱琼瑶[1] 

机构地区：[1]燕山大学信息学院通信与电子工程系,秦皇岛066004

出　　处：《自动化学报》2011年第3期309-315,共7页Acta Automatica Sinica

基　　金：河北省自然科学博士基金（B2004129）资助

摘　　要：为解决现有融合方法不能根据遥感图像的后续处理要求对融合规则进行自适应调整的问题,本文结合模型算子和观测算子的优点构建数据同化系统,提出了基于数据同化和遗传粒子群方法的遥感图像融合算法.该算法将平移不变性小波变换作为模型算子,将对比度金字塔作为观测算子.由后续处理对图像属性指标的依赖程度确定属性指标的权重,用评价指标的加权和构造目标函数,并用遗传粒子群算法优化目标函数,从而获取高清晰度的遥感图像.本文通过遥感全色图像与多光谱图像、红外图像与可见光图像的融合证明了算法的优越性.To solve the problem that the fusion rules could not be adjusted self-adaptively with the available fusion methods according to the following processing requirements of remote sensing （RS） image,this paper builds a data assimilation system by integrating the merit of model operator and observation operator,and presents an RS image fusion scheme based on data assimilation and genetic particle swarm algorithm.The arithmetic takes the plain-move invariant wavelet transform as a model operator,and takes the contrast pyramid as observation operator.Also the scheme ascertains the weights of the attribute indices according to their dependent degrees to attribute indices of image by following processing.And then the algorithm designs the objective function by using the weighted sum of evaluation indices.Moreover,the arithmetic optimizes the objective function by employing genetic particle swarm algorithm to obtain a higher definition RS image.This paper proves the superiority of the proposed scheme by syncretizing RS panchromatic image and multi-spectral image as well as inosculating infrared image and visible light image.

关 键 词：数据同化 遗传算法 粒子群算法 遥感图像融合 

分 类 号：TP75[自动化与计算机技术—检测技术与自动化装置]