In this paper creating semantic maps based on laser terrestrial data is shown. Semantic map is based on transformed geometric data (3D laser range finder) into the data with assigned labels. These labels can help in several applications such as navigation of mobile robot by finding traversable and not traversable regions. Computation of large 3D data sets requires high computational power, therefore we proposed the GPU based (Graphic Processing Unit) implementation to decrease the computational time. As a result we demonstrate the computed semantic map for mobile robot navigation.