Investigation of physical sensor models for orbit modeling

Currently, a number of control points are required in order to achieve accurate geolocation of satellite images. Control points can be generated from existing maps or surveying, or, preferably, from GPS measurements. The requirement of control points increase the cost of satellite mapping, let alone it makes the mapping over inaccessible areas troublesome. This paper investigates the possibilities of modeling an entire imaging strip with control points obtained from a small portion of the strip. We tested physical sensor models that were based on satellite orbit and attitude angles. It was anticipated that orbit modeling needed a sensor model with good accuracy of exterior orientation estimation, rather then the accuracy of bundle adjustment. We implemented sensor models with various parameter sets and checked their accuracy when applied to the scenes on the same orbital strip together with the bundle adjustment accuracy and the accuracy of estimated exterior orientation parameters. Results showed that although the models with good bundle adjustments accuracy did not always good orbit modeling and that the models with simple unknowns could be used for orbit modeling.