How to get a realistic and efficient camera rendering?

[omichel]

I need to improve the quality of the 3D rending in my simulation without sacrificing the speed. Any hints?

[omichel]

The quality of the rendering depends on the Shapes resolution, on the setup of the Materials and on the setup of the Lights.

The bigger the number of vertices is, the slower the simulation is (except obviously in fast mode). A tradeoff has to be found between these two components. To be efficient, Shapes should have a reasonable resolution. If a rule should be given, a Shape shouldn't exceed 1000 vertices. Exporting a Shape from a CAD software generates often meshes having a huge resolution. Reducing them to low poly meshes is recommended.

The rendering is also closely related to the Materials. To set a Material without texture, set only its Appearance node. Then you can play with the diffuseColor field to set its color (avoid to use pure colors, balancing the RGB components gives better results). To set a Material with texture, set only its ImageTexture node. Eventually, the specularColor field can be set to a gray value to set a reflection on the object. The other fields (especially the ambientIntensity and the emissiveColor fields) shouldn't be modified except in specific situations.

The color field of the ElevationGrid shouldn't be used for a realistic rendering because it is not affected by the ambient light with the same way as the other Shapes.

Here is a methodology to set up the lights:

1. Place the lights at the desired places. Often, a single directional light pointing down is sufficient.
2. Set both their ambientIntensity and their intensity fields to 0.
3. Increase the ambientIntensity of the main light. The result will be the appearance of the objects when they are in shadows.
4. Switch on the shadows if required. The shadows are particularly costly, and are strongly related to the Shapes resolution.
5. Increase the intensity of each lamp.