RPC UNCERTAINTY PARAMETERS: GENERATION, APPLICATION, AND EFFECTS

Commercial satellite imagery contains meta-data for the RPC sensor model. In addition to the RPC ground-toimage polynomial, this meta-data contains two uncertainty parameters, termed ErrRand and ErrBias, that are used for error propagation in support of mono and stereo extraction. These parameters describe the uncertainty in the RPC ground-to-image relationship on a per image basis due to errors in the underlying physical sensor model and its meta-data used by the image vendors to generate the RPC ground-to-image polynomial, along with RPC polynomial fit error. The definition and proper application of these uncertainty parameters are not clear throughout the geopositioning community. This paper presents their recommended definition, optimal algorithms for their generation, and optimal algorithms for their use. In addition, it recommends the addition of a priori inter-image and intra-image correlation functions to be published by the image vendors for optimal flexibility and performance. These functions are image independent for images from the same sensor. They account for temporal correlation of RPC errors between same-pass images as well as correlation of RPC errors for two points within the same image. This paper also describes the effects of RPC uncertainty parameters on geopositioning, and in particular, error propagation and accuracy prediction. Various examples are included.