Geo-specific Scene Design

Aerial images are images taken by satellites or drones. They give you the location of streets, buildings, vegetation and if the resolution permits, road markings and stencils. Aerial imagery doesn’t contain any information about object height or the height profile of the roads and the landscape. However, if you combine an aerial image with elevation data, the image will bend to the underlying height profile, which is convenient for placing objects.

Elevation data comes as an image where gray values encode height. These images give you the landscape height profile, but they do not contain any information about the roads or other objects in the scene. Furthermore, the height information is typically given in such a way that buildings, trees and other objects do not appear in the data.

Point cloud data is typically recorded by a lidar and contains a set of 3D points that belong to the scene objects. You can use the point cloud data to get the height profile of roads and surfaces. If the point cloud is dense enough, you can also see the location and height of trees, buildings and other scene objects. However, note that the height profile of roads and surfaces is not smooth, as it contains trees, cars and other objects that were present during the collection of the data. This can be a disadvantage of point clouds with respect to elevation data when adapting road heigh profiles.

Vector data consists of a set of points and geometric figures such as lines and rectangles, each providing information about some object in the scene. Apart from the location of scene elements, this data contains other information such as names of roads and house numbers. However, the representation of lanes is rather simple, so that you may not see details such as carved out turning lanes, for example. Furthermore, this data does not contain any height profile information and does not bend when combined with elevation data, as aerial imagery does.