Impact of spatial correlations on the surface estimation based on terrestrial laser scanning

Abstract

In terms of high precision requested deformation analyses, evaluating laser scan data requires the exact knowledge of the functional and stochastic model. If this is not given, a parameter estimation leads to insufficient results. Simulating a laser scanning scene provides the knowledge of the exact functional model of the surface. Thus, it is possible to investigate the impact of neglecting spatial correlations in the stochastic model. Here, this impact is quantified through statistical analysis. <br> The correlation function, the number of scanning points and the ratio of colored noise in the measurements determine the covariances in the simulated observations. It is shown that even for short correlation lengths of less than 10 cm and a low ratio of colored noise the global test as well as the parameter test are rejected. This indicates a bias and inconsistency in the parameter estimation. These results are transferable to similar tasks of laser scanner based surface approximation.