Insights into the affine model for high-resolution satellite sensor orientation

Abstract

As a model for sensor orientation and 3D geopositioning for high-resolution satellite imagery (HRSI), the affine transformation from object to image space has obvious advantages. Chief among these is that it is a straightforward linear model, comprising only eight parameters, which has been shown to yield sub-pixel geopositioning accuracy when applied to Ikonos stereo imagery. This paper aims to provide further insight into the affine model in order to understand why it performs as well as it does. Initially, the model is compared to counterpart, ‘rigorous’ affine transformation formulations which account for the conversion from a central perspective to affine image. Examination of these rigorous models sheds light on issues such as the effects of terrain and size of area, as well as upon the choice of reference coordinate system and the impact of the adopted scanning mode of the sensor. The results of application of the affine sensor orientation model to four multi-image Ikonos test field configurations are then presented. These illustrate the very high geopositioning accuracy attainable with the affine model, and illustrate that the model is not affected by size of area, but can be influenced to a modest extent by mountainous terrain, the mode of scanning and the choice of object space coordinate system. Above all, the affine model is shown to be both a robust and practical sensor orientation/triangulation model with high metric potential.