Semantic Segmentation of Fused Mobile Mapping Data

To create a digital twin of a city, all relevant objects, within the city, must be mapped and labelled accordingly. In order to cover the entire city, vehicles and aircraft are used for the mapping task. While airborne mapping can cover areas that are impossible for a vehicle to reach, terrestrial mapping provides a different perspective containing further necessary information about the city. The two datasets are registered using reference points, fused together, and processed so that semantic segmentation can be performed by a neural network. The segmented point cloud can then be exported to create a 3D digital twin of the city. This work presents a process for the construction of a 3D digital city model from airborne and terrestrial mapping data.