Abstract. The development of effective solutions for automated vehicles ensuring high levels of safety in any scenario have to be tested in many real world conditions. Large datasets describing real user behaviors in many conditions should be collected to this aim. To this purpose, a quite attractive option is that of using Unmanned Aerial System (UAS) imagery, which can be effectively used to extract real world driver trajectories, and to extract information of interest concerning both their interactions and the context. This work aims at providing an assessment of the accuracy of the geometric information that can be extracted on users’ trajectories and to present a strategy for the estimation of the 3D shape of the involved vehicles. The obtained results show a quite remarkable performance in terms of ability of determining a proper description of the characteristics of the vehicle trajectories, i.e. speed and track.