heal.abstract |
The aim of this paper is to study the impact of using various models for the determination of orientation parameters or georeferencing of high resolution satellite images, on the accuracy of the extracted DSM and orthorectified images. Empirical and physically based models are investigated: Rational Functions (RFC), Toutin's model and orbital parameter model, in which the satellite trajectories can be predicted or simulated by using the physical properties of the satellite orbits. Thus, orientation of every image of the scene can be determined with the Keplerian elements, the location and attitude of the platform. A pair of IKONOS stereo Geo product images, at the north-western part of Athens, Greece, is used for the comparison tests. Twenty three ground points were measured by GPS, scattered all over the area covered by the images. Some of these points have been used for the determination of orientation parameters as GCPs, and the rest as check points. The process for DSM and ortho-image production, were done using: OrthoEngineSE of PCI Geomatics software (using the RFM/RPCs with four different combinations of GCPs and the Toutin's model with 9 GCPs) and Leica Photogrammetry Suite (using RFM with two combinations of GCPs). The orbital model was used, in two data sets: the above IKONOS image pair and a SPOT5 image, in which 39 GSPs were available. Quality and accuracy controls on the ortho-rectified images, which were produced with all the above techniques, were made. The quality control was made by optical inspection of the whole area under study. Accuracy controls included the calculation and statistic analysis of the deviations of the measured on the orthoimages coordinates from the known accurate coordinates of the 14 check points. Also, accuracy controls were made between the ortho-images, which were produced from PCI and LPS, using various combinations of GCPs; the coordinate deviations were of the size of 1.2÷3.6 m. For each software package detailed conclusions are given for the functionality and the easiness in use, the accuracy and the completeness of the produced DSM, the errors of orthoimages and the improvement of accuracy using more control points. |
en |