Investigations on Applicability of Affine Transformation for Geo-Referencing Of Cartosat-1 Imageries

The information content of satellite imageries as a mean for providing usable data for disaster forecasting and management, management of natural sources, remote sensing and photogrammetric applications has attracted the worldwide attentions during the last two decades. Since geometric correction of these images is the first and a crucial step for using them, a great deal of research efforts has already been focused on the subject worldwide. Two main approaches are used for geo-referencing of linear array imageries. The first approach is the so called rigorous model. This approach is based on the physical modeling of the motion and attitude variation of the imaging sensor during the acquisition time and is the most accurate and precise model for geometric correction. This model is based on the so called collinearity condition equations and needs the sensor calibration data and the orbital parameters of the satellite, which is not always accessible due to security and economic restrictions. The second approach, known as generic, is independent of any metadata information and solely relies on the ground control points. Varieties of generic models, from rational functions to simple DLT, are in use. However, the great difficulties with these approaches are their dependence on large number of well distributed GCPs. The most precise generic model which is used as a substitute for rigorous model is rational function model (FRM). It has been shown empirically that RFMs in terrain independent mode, which is calculated by fitting the rational function coefficients to the rigorous model, approaches the accuracy of rigorous model. Nevertheless, a very interesting parametric approach, which uses simple 2D to 3D affine transformation, has been experimentally proven to be very promising. This approach has been already evaluated by different researchers worldwide and reasonably accurate results in both flat and hilly terrains have been reported using only few numbers of GCPs. The theoretical basis that justifies the achieved accuracy is the fact that with the high resolution satellite images the very small camera field of view and the high flying altitude makes the incoming signals almost parallel. This renders the perspective geometry along the scan lines to approach the parallel geometry and effectively a homogenous geometry in the scan line and the direction of the satellite motion is produced. This particular geometry provides a simple linear relationship between the image space and the object space and makes a simple eight parameters affine transformation optimum for geo referencing applications. Simplicity of the formulation (i.e. only eight affine parameters for the entire scene and linear form of the equations), and also few numbers of required GCPs and the achieved accuracy makes this approach very attractive from the mapping point of view. The main intention of this paper is to evaluate the potential of the 3D affine transformation as applied to the Cartosat-1 images by implementing an approach which is independent of the field measurements. This is achieved by fitting the affine transformation parameters to the supplied rational function coefficients. Since this approach is not restricted to the field measurement for generating GCPs, a more realistic evaluation of the nature and the limitations of the affine transformation are achieved by generating well distributed fictitious GCPs. Other related issues such as the earth curvature and the influence of the terrain topography are also discussed and analyzed.