In the recent decades, the determination and evaluation of geometrical correction models as well as georeferencing satellite images have been of great consideration due to their frequent use in various fields, and are regarded a leading topic in photogrammetry and remote sensing. This paper is about the geometric correction of the Worldview-2 satellite image using different modeling methods and tries to give an overall evaluation of strength of various possible modeling for a prototype image of an urban area like Tehran. The distribution and number of control points with regard to their effects in each modeling method were examined which resulted in a high precision of a final geometry correction about 0.36 meter using rational functions. For more optimization artificial intelligent methods like genetic algorithms and neural networks were used. With the use of perceptron network, a result of 0.71 pixels with 4 neurons in middle layer was gained and the final conclusion was that with these algorithms it is possible to optimize the existing models and have better results than usual ones.