论文部分内容阅读
多源遥感数据分析应用前需要评价影像几何校正或配准的质量以确保满足应用的需求。传统的均方根误差(RMSE)位置精度评价方法固然简单,然而无法描述校正后控制点(GCP)残差的分布特性,因而具有一定局限性。采用交叉验证法并从GCP残差角度出发,分别引入了Moran’sI空间自相关系数和标准偏差椭圆用于评价GCP残差的随机性和方向性。模拟实验结果表明,Mo-ran’sI和标准偏差椭圆可用于定量衡量GCP残差的空间随机性和分布方向性,从而更深入地分析几何校正的效果,指导选择恰当的校正模型,提高影像校正的精度。
Before applying multi-source remote sensing data analysis, the quality of image geometric correction or registration needs to be evaluated to ensure that the application needs are met. The traditional root mean square error (RMSE) position accuracy evaluation method is simple, but it is unable to describe the distribution characteristics of the corrected control point (GCP) residuals, and therefore has some limitations. Using the cross-validation method and the GCP residuals, Moran’s spatial autocorrelation coefficient and standard deviation ellipses were introduced respectively to evaluate the randomness and directionality of GCP residuals. The simulation results show that Mo-ran’s and standard deviation ellipses can be used to quantitatively measure the spatial randomness and distribution direction of GCP residuals, so as to further analyze the effect of geometric correction and guide the selection of appropriate correction models to improve image correction The accuracy.