Abstract: Cloud maps are the most frequently acquired sources nowadays because they contain very useful meteorological information. So the cloud maps are used as input to drive the simulation of 3D clouds. Using gray level co-occurrence matrix and fractal theory, the texture feature of cloud maps are extracted, basic cloud types are classified, and so the distribution feature of cloud scene is retrieved. The coarse distribution is mapped to the surface of the earth to produce basic parameters of atmosphere. With these basic parameters of atmosphere, CML method is used to generate 3D cloud density distribution, then 3D clouds are rendered onto the spherical surface cover the earth. The auto analysis of Fengyun2 cloud maps and visualization of 3D cloud scene on digital earth are realized.

Key words: computer application, visualization of clouds scene, driven by cloud-map data, texture features extraction