Automati̇c object extraction usi̇ng terrestrial laser scanner data

Many study area such as GIS applications, cultural heritage documentation and urban planning applications needs 3D buildings models. The required data for 3D model generation can be obtained rapidly and accurately with terrestrial laser scanner. Nowadays, Data acquisition process with terrestrial laser scanner could be carried out automatically with little user response. On the other hand, Automatic 3D model which was generated using terrestrial laser scanner data, is a research subject. Generally, there are two different approach for automatic 3 model generation using terrestrial laser scanner data. These are meshing and geometric primitives modelling. Meshing is methot that is used for complex surfaces that are hard to define mathematically. Geometric primitives modelling is an approach that is used for simple surfaces which are easy to define mathematically. The most important step of geometric primitives modelling is segmentation that is done classification and clustering of data. In this thesis study, segmentation step of geometric primitives modelling approach is carried out. Within this scope, Random Sample Consensus (RANSAC) algorithm was applied on point cloud data that is obtained from terrestrial laser scanner. Then, objects that has planar, cylindrical, conical and spherical surface was automatically extracted in point cloud data. Automatic extracted surfaces was compared with manual extracted surface for the evaluation of accuracy of RANSAC algorithm on surface extraction. Obtained results shows that RANSAC algorithm is effective tool for automatic extraction of objects which have simple geometric shapes. Keywords: Terrestrial Laser Scanner, surface extraction, RANSAC Algorithm, Segmentation, 3D modelling