We present an efficient method for creating large-scale animations of vertical developing cumulus-type cloud growth. The dynamics of cloud formation, growth, and motion are complex phenomena, and depicting these dynamics remains a significant challenge in the area of simulation and animation of natural phenomena in computer graphics. A novel aspect of this paper is the combination of a physical simulation method and a stochastic simulation method for obtaining an animation of large-scale phenomena with effects that are more natural. The physical simulation method is used to accurately solve fluid dynamics on a relatively small scale and prepare a 3D primitive pattern of a realistic animation of cloud growth. The stochastic simulation method uses 1/f^β noise functions and performs a large-scale simulation of an air current caused by the rising and condensation of water vapor due to the thermal effect. Many copies of the 3D primitive pattern are recursively mapped into the air current to constitute a large-scale continuous cloud growth. This combination of the physical and stochastic simulation methods can animate large-scale phenomena efficiently without enormous computational time and memory. The physical and stochastic simulations are achieved by particle-based methods, and the mapping is applied on simulated particles. Experimental results show that the proposed method efficiently creates realistic animations of large-scale cumulus and cumulonimbus clouds.

View full abstract

Registration method unifies some point clouds measured by range sensor in order to get geometric information of whole building. In our previous method we use feature-line matching approach to get geometric information of a building but the result is not good enough because of the limited number of feature lines. In this paper, we extract ridge line to increase the number of feature lines. First, we applied the hierarchical crusting method which based on flatness of building to measured point clouds, then made a boundary line of adjacent two regions as a ridge line. As a result, our extended method has improved the restoring building and the strong constraint measurement conditions, compared to previous method.

View full abstract

We developed a simple and low cost wind velocity sensor based on a small microphone. In this paper, we first present the mechanism of the sensor, i.e., how to find the wind velocity from the output wind noise of a microphone by successfully suppressing the influence of other sounds existing in the environment. Then, we demonstrate the effectiveness of the sensor by showing the application to the real-time animation of a tree swaying in real wind.

View full abstract

[in Japanese]

View full abstract