基于改进的自动散点控制回归算法的遥感影像相对辐射归一化

多时相遥感影像的相对辐射归一化是进行变化检测或拼接不可缺少的步骤。针对现有方法的不足,以自动散点控制回归(Automatic scattergram-controlled regression,ASCR)技术为基础,提出了一种改进的ASCR算法(Im-proved automatic scattergram-controlled regression,IASCR)。它的核心思想是首先利用粗剔除和主成分分析的方法选择占主体信息量的“未变化”像元,然后再利用最小二乘法确定回归方程,对多时相遥感影像进行归一化。以北京地区的多时相TM影像为实验数据对其进行归一化,并与ASCR法、全景简单线性回归法(Simple regression)、暗-亮(Dark setbright set)归一化法、伪不变特征归一化法(Pseudo-invariant feature)的结果进行了比较。实验结果表明,IASCR算法是解决多时相遥感影像辐射归一化问题的有效手段。

Relative radiometric normalization of remotely sensed images based on improved automatic scattergram-controlled regression

Relative radiometric normalization of multi-temporal satellite images is the necessary process in change detection or image mosaic.An improved automatic scattergram-controlled regression technique(IASCR) is developed to solve the shortage of current methods.The main idea of the new method is as follows: firstly the majority of "no-change" statistic pixels are chosen through coarse-eliminating and principal component analysis,then the regression coefficients are computed from least-squares equations,and at last,the multi-date satellite images are normalized using coefficients computed.Some Beijing multi-temporal TM images are used to assess the IASCR,and some experiments are made to compare the new method with other methods such as ASCR normalization,simple regression normalization,dark set-bright set normalization,pseudo-invariant feature normalization.The results show that the IASCR is an efficient approach to solve the problems of multi-date satellite images normalization.