Thesis - Real Time Snow Rendering
Driving vehicles on roads in winter conditions can be extremely dangerous at times. Winter snow and fog causes problems while driving. When snow is blowing around, it is hard to see other vehicles or judge their speed. Detecting the changes in speed of the vehicle or snowplow in front of you is important for safe driving. Blowing snow and fog complicates visibility while driving. Due to the extreme weather conditions and poor visibility, it becomes difficult to detect speed changes of other vehicles. The poor visibility affects our view, resulting in the potential for rear end collisions.
We can create a 3-D simulation of the extreme winter conditions where a person can drive around or follow other vehicles or snowplows in blowing snow or foggy conditions. The interactive virtual environment can help us in studying user reaction times to the changes in speed/motion of other vehicles or snowplows. We could test the user reaction times for detecting motion of the snowplow for a variety of color and lighting configurations. Determining the optimal color and lighting configurations for snowplows could help in making driving safe in extreme winter conditions. For this virtual environment to be effective, the rendering has to be in real time. Simulating effective snow in real time is a challenging task as there are no good solutions available for real time rendering of snow.
The aim of my research is to improve the visual rendering of snow taking into account the physical interaction between light and the snow particles. The methodology that has been used till now to produce realistic effects of light-object interaction is ray tracing. Ray tracing has been used for rendering images ahead of time by tracing the path of light through every pixel in the image. Due to the high computational costs involved, ray tracing is not suitable for rendering interactive environments. We need a method that could approximate the ray tracing effects by reducing the computational costs and give us acceptable and realistic results.
Our methodology for simulating snow in a real time environment is based on rendering large number of dynamic snow particles. We explicitly model the blowing snow and its effect on the environment. The lighting calculations are applied to each snow particle to characterize the absorption, emission and scattering of light. The presence of dynamic lights in the scene affect the lighting calculations for the snow particles giving us more realistic scenes encompassing various optical effects like scattering, high dynamic range rendering, and blurring.
We use imposters for our particle system which was developed previously. I have modeled the snow particles using snowflake shapes and Gaussian function for transparency. The lighting calculations for every snow particle are done using the Henyey-Greenstein scattering distribution model. We have also implemented High Dynamic Range rendering to give more realistic effects to the scene. We are using additive alpha blending for the transparent snow particles which might have to be changed in the future for effects like shadowing. We have also incorporated fog effects which makes the objects at distance blurry depending on the blowing snow density. Our 3-D environment for snow is very fast and works in real time giving us realistic results of snowing and foggy conditions.
Downloads
Videos