Systematic error compensation for RPC model using semi-parametric estimation

When the conventional method of parameter estimation is used to construct the RPC model, the uncertain factors can result in inconsistency between the RPC model and the reality. The inconsistency shows up as evident systematic error in the constructed RPC model. To tackle the problem, a non-parametric component is introduced on the base of the parametric model to account for the unknown factors and their effects. The new method, namely the semi-parametric estimation, could effectively compensate the effect of systematic errors. This paper studied the construction of the RPC model of remote sensing images using the semi-parametric estimation method. The experiment with SPOT-5 imagery demonstrated that the semi-parametric estimation method could improve the precision of fitting the rigorous imaging model by RPC model.