Creation of the accurate raster driven polygonal environment for the 3D surface models based on the LIDAR technology

Abstract

The paper describes the new way of creation of 3D model of the surface, which uses the raw LIDAR point clouds as an input. The newly proposed way is designed in the form of the parametrical and procedural model which unifies the following partial models together: terrain, vegetation, clouds and buildings. All the partial models are derived from the input LIDAR data, except for clouds. Model of clouds is derived from NOAA data. The particular issue mainly addressed by this article is vegetation. Vegetation is represented by millions of individual points in the LIDAR scan. It results into high requirements on the computer hardware, especially the memory. Thus, authors propose a way how to convert vegetation point clouds into the quadrilateral polygonal form. Then, quad shaped objects are scattered on the top of the terrain model based on the exact LIDAR attributes. A height and position of each individual tree is another significant issue targeted by the proposed model. These two parameters are coded in the black-white raster. This raster works as a parametrical mask for each object. Next, proxy representatives of vegetation are created. The same approach is applied in the case of weather conditions. Resulting 3D models are generated independently on the very large LIDAR point clouds. Their attributes are derived from the 8 or 16 bit rasters. Vegetation is rendered only when it is needed. Its visualization is done by means of proxy representatives. The proposed approach allows creation of photorealistic and scientific 3D models of the surface for the very large areas of interest. At the same time, this procedure preserves the smallest real-world details like exact height of each tree.