Orthorectification of Bilsat Imagery Using Rigorous and Simple Geometric Models

The orthorectification accuracies of Bilsat Multispectral (XS) imagery using a rigorous and eight simple geometric models are presented. Majority of the selected test area was covered by the existing large scale digital orthophotos, which were used as the source to collect Ground Control Points (GCPs). For the orthorectification process, a DEM generated from stereo ASTER images was used. The developed rigorous model utilizes the well known collinearity condition equations to model the physical reality of the Bilsat sensor. The initial attitude and position information were obtained from the given ephemeris data and the refinement was carried out using an iterative least squares adjustment procedure. The simple geometric models used include the 2D polynomials (1, 2 and 3 order), Rational Functions (1, 2 and 3 order), Projective Transformation (1 order), and Direct Linear Transformation (DLT) (1 order) models. It was found that, using 4 GCPs and 40 Independent Check Points (ICPs), the rigorous model produced RMS errors of 0.75 and 1.14 pixels respectively, for GCPs and ICPs. It was also found that, if a reasonable number of GCPs are used, the rigorous model provides a RMS error of around one pixel for ICPs. Among the simple geometric models used, the 1 order 2D Polynomial model provided the highest RMS error of 3 pixels. On the other hand, the DLT model produced the lowest RMS error of 0.61 and 1.09 pixels for GCPs and ICPs, respectively.