significant part of the data points obtained by using airborne laser scanning technology come from points reflected from objects situated above the ground such as trees, shrubs or buildings. Clear-cut and accurate segmentation is a crucial stage in data processing which allows to identify the homologous regions in terms of specific properties within a dataset of points, which further allows to generate DTM's or model building shapes. This paper shows an analysis of the two most commonly used algorithms for ALS point cloud filtering: active TIN model and linear prediction. The study was performed on 24 specifically extracted testing samples characterized by different topography and land use. The verification of the results of the automatic filtration process of both algorithms was based on comparison to reference datasets. As a result of this comparisons the relative percentage errors of automatic segmentation were determined. The level of the estimated errors varies from 0% to around 20% and depends on the characteristics of the land and the objects which are on the surface. The conducted study confirmed the high efficiency of both evaluated algorithms, at the same time revealing their limitations and differences in the filtration process for areas of a complex topography or terrain coverage. Both algorithms provide similar classification of point clouds describing land use for agriculture, areas on which a single building, shrub or tree is located, and for used car parks. Method based on linear prediction works better than active algorithm TIN model in terms of points recorded by the laser beam being reflected from vehicles/flyovers/bridges.a