Fitting using Kriging, originally developed for geological exploitation, has here been applied for fitting an expected pattern to noisy, irregular, in-flight measurements of a satellite antenna. The noise level in in-flight measurements is often so high that only the central part of the main beam appears. Using the Kriging method, a characteristic function - the regression model - was first fitted to the measurements. For the main beam, this was chosen to be described by a general second-order polynomial. To this was added a more-detailed correlation model, which represented realistic deviations from the regression model but filtered out the fast variations of the noise. The method was applied to simulated measurements of the Planck RF telescope. The results presented showed a considerable reduction of the noise floor of the pattern: even beam details invisible in the original measurements (a shoulder) were revealed by the pattern fitting.