Collecting spatial data using laser scanning is developing quickly. Point clouds, which are primary result of laser scanning, are useful for 3D modelling of scanned objects and Earth's surfaces. Because point clouds are useful in numerous fields and provide a large amount of data, there is a need to automatically extract certain features from point clouds. This thesis describes point cloud segmentation using Gaussian sphere. We used this method on actual terrestrial scanned test data. We described the process in steps from obtaining test data to final results
of segmentation. We focused on three geometrical objects, that are plane, cylinder and cone. We also described basic theory about the Hough transform, binarization and changing test data in discrete form, that were used in the process of segmentation. For processing the test data, we wrote a program in Python. With Python and its add-on libraries we plotted intermediate and final results, which we commented and evaluated. Final thoughts about the described segmentation method are given in conclusion.