A broad range of scientific, administrative, industrial and private users demand for detailed three dimensional (3D) spatial data with typical applications in, for instance, as-built documentation, construction engineering, navigation applications and forensic investigations. The acquisition of 3D spatial data is conveniently performed by means of laser range sensors which provide immediately numerous 3D object points. These 3D point clouds are obtained in the local coordinate system of the lidar sensor which is also known as laser scanner. In order to capture a complete view of objects, multiple sites are necessary. Aforementioned applications frequently demand, besides a common local coordinate system, the transformation into a global coordinate system (geo-referencing). For a time-saving and efficient provision of geo-referenced 3D point clouds, sophisticated methodologies are required to obtain the parameters for the transformation. The main objective of this thesis is the development of an efficient methodology for the determination of parameters and its uncertainty measures for the direct geo-referencing of 3D point clouds acquired by stationary laser scanners. This objective is addressed by a prototypic mss and a novel algorithm to perform the geo-referencing of 3D point clouds considering uncertainty measures. The developed mss is established by the fusion of a laser scanner and gnss equipment. The use of only physically attached sensors to the laser scanner and the waiver of auxiliary control points leads to a time-saving, simultaneous acquisition of 3D point clouds and its geo-reference information, i.e direct geo-referencing. The novel algorithm is fundamentally based on the motion modelling of the mss in an ekf. The circular motion of the laser scanner about its vertical axis is exploited to obtain time series of 3D positions by means of gnss equipment eccentrically installed on the laser scanner. These time series enable a redundant estimation of the transformation parameters and their variance matrices which increase the fidelity and reliability of the proposed approach. The prediction and filtering of the 3D trajectories are realised by two distinct variants of modelling the system. First, states and adaptive parameters are considered in a single antenna scenario. Here, the benefit can be stated by the verification possibility of mss specific parameters which are usually obtained by separate calibration procedures. Second, constraints among the states are taken into account by a dual antennas scenario. The constant velocity of the circular motion and the distance between the two gnss antennas are considered as constraints. The geo-referenced 3D point clouds are supplemented with uncertainties induced by the transformation parameters (transformational uncertainty) as well as the laser scanner (positional uncertainty). Both uncertainties are jointly propagated to the geo-referenced 3D point clouds. The result is a geo-referenced stochastic 3D point cloud which provides spatial and supplemental stochastic information. In addition, an optimisation of transformation parameters by an icp based matching (Gref-ICPHe3) is proposed which utilises both kinds of uncertainties and also incorporates the uncertainties of the relative transformation parameters within the iterative matching. The potential of the developed methodology is shown in practical experiments for both system modelling approaches (single and dual antennas scenario). The results show good coincidence of overlapping regions from different sites and differences of only a few centimetres with respect to known control point coordinates. Owing to the use of one set of transformation parameters for the geo-referencing of the entire 3D point cloud, each 3D point cloud is consistent. It can be stated that the main influencing factor of the proposed geo-referencing approach is the heading and its uncertainty. As expected, the heading and its uncertainty is strongly related to the baseline length used for the heading determination. The proposed methodology with the introduced mss and novel algorithm is suitable for the above-mentioned laser scanning applications. In addition, optimal initial values with uncertainty information for matching algorithms can be obtained. The provision of the geo-referenced stochastic 3D point clouds is efficient and paves the way for rigorous consideration of the entire uncertainty budget of a 3D point cloud also for subsequent analysis like, eg estimation of geometric primitives.