In last years, accurate spatial data from high resolution satellite images are getting more and more frequently used for modelling topography and other surveying purposes. To extract accurate spatial information, a sensor's mathematical models are needed. Those models classified to two branches: rigorous (parameirical or physical) models and non-rigorous models. In the paper a dynamic sensor model is proposed to extract spatial information from geo-rectified images named the geo-images which their geometry at the time of imaging have been lost. The developed model has been reconstructed basing on a transformation of central-perspective projection into a parallel one.

Nowadays, an orthomap destined for different purposes can be created from High Resolution Satellite (HRS) images using IKONOS, QuickBird and other satellite imageries having Ground Sampling Distance (GSD) lower than I m. The orthomap is one of the main sources for establishing GIS. Accuracy of the orthomap depends first of all on the parameters of Ground Control Points (GCPs) (the forms, number, accuracy and their distribution). In order to reduce the cost and number of GCP field measurements, the block of HRS images has been proposed. The accuracies of determined points in the block of HRS images are affected by the mathematical model used to build a block. The paper presents a general algorithm of bundle block adjustment model of HRS images using Keplerian parameters. In order to overcome strong correlation among exterior orientation elements of HRS images that causes the normal equation ill-conditioned, the ridge-stein estimator and orbital addition constraints have been proposed.