Generating Image Sequences of Augmented Road Scenarios

The traffic scenario consists of the road scenario and traffic elements. The original image sequences can be captured by a moving platform equipped with camera by driving along the road. In this paper, an approach based on augmented reality is proposed that supplements new elements (traffic cones, obstacles, etc.) into those image sequences. We model the traffic scenario in the 3D space of the scene, and all the traffic element models are unified in this 3D space. 3D registration is employed to compute the projection positions of inserted traffic elements. The normalized viewpoint models are used to estimate the projection appearance. Depth is used to determine the projection order. The experimental results verify that the proposed approach can generate various new image sequences of traffic scenarios, which can be provided as benchmark data for the simulation testing of unmanned vehicles.