The main idea of this project is to introduce a conception of semi - automated method for building model extraction from Airborne Laser Scanning data. The presented method is based on the RANSAC algorithm, which provides automatic collection planes for roofs model creation. In the case of Airborne Laser Scanning, the algorithm can process point clouds influenced with noise and erroneous measurement (gross errors). The RANSAC algorithm is based on the iterative processing of a set of points in order to estimate the geometric model. Research of u sing algorithm for ALS data was performed in available Cloud Compare and SketchUP software. An important aspect in this research was algorithm parameters selection, which was made on the basis of characteristics of point cloud and scanned objects. Analysis showed that the accuracy of plane extraction with RANSAC algorithm does not exceed 20 centimeters for point clouds of density 4 pts . /m 2 . RANSAC can be successfully used in buildings modelling based on ALS data. Roofs created by the presented method could be used in visualizations on a much better level than Level of Detail 2 by CityGML standard. If model is textured it can represent LoD3 standard.