The Application of Data Fitting in Direct Geolocation of Remote Sensing Data

In direct geolocation of remote sensing data, the high precise ephemeris and attitude data is required. Due to the measurement error, as well as the discreteness of the ephemeris data, it is necessary to fit the ephemeris data, before direct geolocation is applied. Commom algorithms for data fitting include Lagrange interpolation, polynomial fitting, cubic spline fitting and cubic uniform B-spline fitting. After a short introduction to the mathematical model of these algorithms, a systematic comparison of the performance among them is conducted on a full scene SPOT image obtained. The experimental results show that, for mountainous terrain, Lagrange interpolation outperforms the other three algorithms, while for flat terrain, cubic uniform B-spline fitting achieves the best performance.