遥感影像位置不确定性建模的多点地统计模拟方法  

Multiple-point Geostatistical Simulation for Modeling Positional Uncertainty in Remotely Sensed Images

作　　者：王剑旭 张景雄[1] 贺超 WANG Jianxu;ZHANG Jingxiong;HE Chao(School of Geodesy and Geomatics,Wuhan University,Wuhan 430079,China;Aerospace Science and Industry Intelligent Operation Research and Information Security Research Institute(Wuhan)Co.,Ltd.,Wuhan 430048,China)

机构地区：[1]武汉大学测绘学院,湖北武汉430079 [2]航天科工智能运筹与信息安全研究院(武汉)有限公司,湖北武汉430048

出　　处：《测绘与空间地理信息》2024年第10期16-19,22,共5页Geomatics & Spatial Information Technology

基　　金：国家自然科学基金资助项目——上海多源卫星遥感数据高精度几何校正研究(210700079)资助。

摘　　要：为了对遥感影像的位置不确定性进行建模分析,本文提出了一种顾及趋势面的多点地统计模拟方法。此方法对研究区的影像位置误差趋势面进行拟合,利用样本点位置误差的残差生成训练图像(TI),然后基于残差的TI对研究区进行直接采样(DS)模拟,从而实现位置不确定性建模。实验结果表明,与条件模拟方式相比,非条件模拟更适合位置不确定性的随机模拟;当位置误差存在明显的整体趋势时,考虑趋势面的非条件模拟方法具有更好的模拟效果。所提方法也适用于其他存在整体趋势的地理空间变量的不确定性建模。In order to model and analyze the positional uncertainty of remotely sensed images,a multiple-point geostatistical simulation method considering the trend surface is proposed.Firstly,the trend surface of image positional errors in the study area is fitted.Then TI of residuals is generated by using the residual of study area,and direct sampling(DS)simulation is carried out for the study area based on TI of residuals,thus using random simulation to achieve positional uncertainty modeling.Experimental results show that compared with conditional simulation,unconditional simulation is more suitable for random simulation with positional uncertainty;when the positional error has an obvious overall trend,the unconditional simulation method considering trend surface has a better simulation effect.The proposed method is also applicable to the uncertainty modeling of other geospatial variables with overall trends.

关 键 词：位置不确定性 趋势面 多点地统计学 直接采样 非条件模拟 

分 类 号：P237[天文地球—摄影测量与遥感]