STOCHASTIC VERSUS NON-STOCHASTIC TREATMENT OF GCPs FOR IKONOS IMAGE AFFINE TRANSFORMATION

Very High Resolution Satellite VHRS image technology is considered as a basic information source for mapping, geographic information systems, and different applications in geomatics. In this paper an IKONOS image covering the area of Rodah Island in Cairo is used and an affine 2D transformation model is utilized for image to ground transformation. Most of the satellite image providers withhold the camera model as well as the precise satellite ephemeris required for the geometric correction of these images. IKONOS imagery is no exception and accordingly we should use empirical methods for image rectification. In this paper we test the performance of the resulting accuracies by adopting two approaches in dealing with the ground control points, namely a) as non-stochastic quantities i.e. weight equals infinity; b) as stochastic quantities i.e. have certain weight. An application using Matlab programming language based on the principle of least squares adjustment was developed to calculate the six affine transformation parameters followed by calculating the positional residuals ∆X, ∆Y, ∆P at both the ground control and check points. Generally, the positional residuals as well as the standard deviations were almost 20% less at the ground control points compared with those at the ground check points. Moreover, the positional residuals and the standard deviations were approximately 15% less when treating the ground control points as stochastic variables versus non-stochastic variables.