Laser scanning data processing is widely used to solve regional planning problems in a GIS environment including Digital Terrain Models (DTMs) analysis and ground surface reconstruction. Some gaps in algorithms for processing of raw laser scanning data during DTM creation are analyzed. Algorithms for filtration, triangulation and defragmentation of laser scanning point clouds are proposed. Advantages and disadvantages of the algorithms proposed are discussed. The proposed triangulation algorithm is used for defragmentation of laser scanning point clouds into semantic component parts. Defragmentation includes recognition of engineering objects and other objects of the terrain, and their delineation. The results of real problems’ solutions described in the paper show the robustness of the proposed algorithms