Automatic Generation of 3D City Models and Depth Videos for Augmented Virtual Environments
Summary
The NLR developed a simulator which allows users to experience aircraft fly-overs at a given location. This is done by adding virtual airplanes to a recorded panoramic video. They want to extend the use of their simulator to demonstrate the impact of future air traffic of drones. In order to correctly render drones behind and in front of objects in the video, depth information is needed. A method was designed and implemented that automatically generates a 3D city model and a depth video, given a panoramic video, estimated GPS coordinates, estimated camera height and publicly available aerial LiDAR data. The 3D city model and depth video are then used to render occlusions of drones from objects in the video. Furthermore, the drones also receive shadows from the 3D city model and cast shadows on it. A subjective study and user study were conducted to evaluate our method. The results showed that our method is able to generate a 3D city model and align it with a panoramic video with only a few errors. However, the generation of a depth video needs much improvement. Furthermore, the results of the user study showed that the realism of the simulations created with our method is satisfactory.