An optimized orbital parameters model for geometric correction of space images

The Orbital Parameters Model (OPM) is a physically constrained model for geometrical correction of satellite images with dynamic geometry. In this model, Keplerian parameters are used to establish an exact relationship between image space and object space. However, insufficient information regarding the precision of the observed Keplerian elements requires certain complementary parameters to be imposed on the model. These parameters lead to an undesirable increase in the number of required Ground Control Points (GCPs). To overcome this problem, unknown parameters may be introduced as quasi-observations in the adjustment procedure. This paper demonstrates that quasi-observation weights have a large impact on the OPM solution. For the first time, a Variance Components Estimation (VCE) technique is used to optimize the weights of quasi-observations when information on their observed precision is lacking. The proposed method is evaluated on two different SPOT datasets and one RapidEye dataset of stereo images. The results demonstrate the efficiency of VCE-based OPM where the highest expected accuracy of 3D information is reached even with three or four GCPs.