Drone cinematography and the generation of environment models for flight planning

Student thesis: Doctoral ThesisDoctor of Philosophy (PhD)


The use of drones (UAV's) as a camera platform has become widespread in media production, finding application in a diverse range of genres such as natural history, sport, news and movies. This thesis examines various aspects of their use for the coverage of live events such as sports.
The techniques and heuristics used in standard cinematography that are still applicable when filming from a drone are reviewed. Current drone cinematography practice is examined and its limitations for live filming are discussed. A novel shot hierarchy and taxonomy suitable for directing and controlling a multiple drone platform to film sports is proposed. A methodology to determine suitable drone camera shot parameters (e.g. drone height or speed), using the subjective testing of simulated camera shots (created using Unreal Engine), is discussed and the results obtained from testing a representative set of the proposed shot types are given.
For successful filming, preparation using a flight planning or training application is an important requirement and for this to be effective the software should incorporate a realistic model of the environment at the filming location. A method of creating these 3D models using photogrammetry, with image data obtained from existing resources such as Google Earth, is investigated. The theory behind the photogrammetry reconstruction process is discussed. In cases where images for photogrammetry are not available, they can be generated using a drone scan of the filming location. An outline of current state of the art research in the optimization of image capture scans for photogrammetry is given.
A system built using the Python development environment in Blender, designed to optimize scanning parameters given a basic 3D model of an environment, is described. The system produces a metric value for the expected quality of photogrammetric reconstruction from a scan by calculating the accumulated coverage of surface points over all images. The approach is evaluated to determine the variation in the reconstruction metric with scan parameters and the results compared to those obtained using simulated scans created in Unreal Engine. The optimum height and camera angle calculated using the system were established to be largely independent of other parameters such as the focal length and the track separation distance. Using a methodology to adjust the parameters of the coverage model to take into account photogrammetry image overlap requirements, it has been found possible to calibrate the system so that the calculated optimum height and camera angles were comparable with the actual results obtained through photogrammetry (using scans simulated in Unreal Engine). The developed system provides significant benefits for the optimization of photogrammetry for 3D environment model creation.
Finally, conclusions are given on the benefits of simulation for camera shot design, and the use and optimization of photogrammetry for 3D environment model creation.
Date of Award22 Mar 2022
Original languageEnglish
Awarding Institution
  • The University of Bristol
SupervisorDavid R Bull (Supervisor) & Andrew Calway (Supervisor)


  • Drone Cinematography
  • Camera shot optimization
  • Camera shot simulation
  • Subjective testing
  • Real World Environment Modelling
  • Photogrammetry optimization

Cite this