The three-dimensional modelling of forest landscape scenes based on laser (LiDAR) scanning data isn’t a trivial task. It requires a getting some subtasks. Firstly starting point cloud clustered into several groups: the point of the Earth, plant mass and man-made objects. Based on the received subset of points it is restored forest vegetation cover in three dimensional space. There are several approaches modeling of complex natural objects of vegetation, which describe using graphical primitives is not possible. The first approach is based on the use of L-systems, the second approach used voxel modeling. Voxel is an extension of the pixel for three-dimensional space and a minimum three-dimensional unit, the smaller the dimension, the more accurate and realistic models of objects. Widespread use of voxel modeling obtained in medicine and computer games. Description of the proposed method is divided into four stages: preparation of the starting point clouds, including the standard data pre-filtering of noise and alignment of data in a single coordinate system; the construction of three-dimensional map of the heights of the original data; display of the results; optimization and visualization of the final model of the landscape.