Stokastik yüzey örgüsü oluşturulması

This thesis aims at developing a novel methodology for 3D surface mesh generation from Terrestrial Laser Scanners (TLS)-derived point clouds. The methodology has been divided into two main parts. The first part deals with developing a generic and anisotropic point error model, and the following second part is focused on the surface (mesh) reconstruction. The point error model is capable of computing the magnitude and direction of a priori random errors with regard to direct TLS measurements (vertical and horizontal angle and the range), and it is described in the form of error ellipsoids for each point. A static and repetitive experiment was designed to determine angular precision values. For the determination of the a priori precision of the range observations, a test stand was designed and measured. All the experiments were conducted using three TLSs, namely the Faro Focus 3D X330, the Riegl VZ400 and the Z+F 5010X. Using the test stand measurements, a priori precision for range measurements was modelled as a function of sensor-to-object distance, incidence angle, and surface reflectivity parameters. The error ellipsoid of each point was then computed using the ABSTRACT (continued) law of variance-covariance propagation. The validation procedure showed that the proposed point error model works successfully. The mesh reconstruction method exploits the previously determined error ellipsoids by computing a point-wise quality measure, which takes into account the semi-diagonal axis length of the error ellipsoid. The co-registered point cloud was voxelized, and then managed by the octree data structure. The points with the best quality were selected from the voxelized data in order to use the best data in the subsequent surface mesh reconstruction step. In order to keep the surface resolution versus completeness in appropriate limits, a threshold value was defined for low quality points' elimination. The 3D surface meshes generated from the proposed model outperformed the ones generated from the original point clouds.