I am trying to create a game where the user scans the room using RGBD camera and reconstructs 3D scene. This scene would be integrated into physics engine, after an object recognition algorithm tells it what objects are there in the scene and of what shape and size. Finally, after being rendered in Physics Engine, user should be able to interact with the environment. I know how to do 3D scene reconstruction from images, I can figure out how to contruct objects and interact with them in Physics Engine. I know how to do object recognition, shape and size. However, I am not sure how to integrate the output of 3D scene reconstruction, Object Recognition as input to the physics engine. Any ideas would be helpful.