Ground Truthing UAV Data for Coastal Applications

Undergraduate Project Thesis
Jonathan Clancy
School of Civil & Environmental Engineering
University of New South Wales
Supervised by Dr. Craig Roberts

Figure 1: Image captured by the eBee Sensefly UAV.



The Australian Coastline is a valuable asset and therefore must be monitored to enable early detection of erosion and coastal variability. Unmanned Aerial Vehicles (UAVs) have the potential to provide an efficient and cost-effective survey technique to map beach terrain for monitoring the changes to the volume of sand at the beach. Current surveying techniques include photogrammetry, LiDAR and RTK-GPS technologies. This project aims to investigate the current survey practice of the Water Research Laboratory (WRL) from the University of New South Wales (UNSW) in monitoring the Collaroy-Narrabeen embayment in Sydney. It also aims to determine whether or not it is feasible to use UAV technology for this application and if it has the potential to replace current surveying techniques.


This network was designed to provide periodic data for coastal monitoring purposes.  By utilising various sites around the coastline of Australia, people can gain a better understanding of the relationship between climate and coastal change. This idea is known as the baseline coastal monitoring concept. The establishment of this network is subject to a three-year experiment on the New South Wales (NSW) coast whereby 10 coastal reference sites have been used to trial the baseline coastal monitoring concept. Figure 1 shows the locations of these sites.

The locations of the coastal reference sites are as follows:

Figure 2: Locations of NSW coastal reference sites.

Current Technologies used for coastal mapping:    
  • GNSS Positioning Techniques (e.g. Real-Time Kinematic GPS)
  • LiDAR
  • Photogrammetry

  • Proposed Technology:    
  • Unmanned Aircraft Systems (UASs)

  • A UAS consists of:
  • Unmanned Aerial Vehicle (UAV)
  • on-board sensors (i.e. a GPS navigation chip, an attitude sensor, a barometer, a radio transmitter and an autopilot circuit board)
  • ground control
  • data links

  • UAV Survey

    A UAV survey was conducted at Narrabeen beach on August 22, 2013 using the eBee SenseFly UAV. This is a small-scale UAV survey that was used to test the performance of a small UAV in coastal conditions to determine whether it is feasible to use for on-going monitoring of coastal erosion. Connection to CORSnet-NSW was used to carry out an RTK-GPS survey as a way of ground-truthing the UAV data. This was done by measuring check points in addition to GCPs with a Leica GPS1200 reciever. A total of 10 GCPs were used as part of the survey.

    Regulations from the Civil Aviation Safety Authority (CASA) must be followed for the safe operation of a UAV.

    Figure 3: Project work flow for a UAV survey.

    Figure 4: Geometry and spacing of ground control.

    Design flight path
    Executed flight
    Flight altitude (m)
    ~ 100
    Lateral image overlap (%)
    ~ 90
    Longitudinal image overlap (%)
    ~ 70
    Ground resolution (cm/pixel)
    ~ 3
    ~ 3
    Flight time (min)
    ~ 24
    ~ 36
    Spatial coverage (ha)
    ~ 14
    ~ 18

    eBee Sensefly Specifications

    Net Weight    
    Camera resolution 
    Maximum flight time  
    Cruise speed
    10-16m/s (36-57km/h)
    Radio link    
    Wind Tolerance
    45km/h (12m/s)

    Figure 5: eBee Sensefly UAV.


    The data collected from the UAV survey was post-processed using Pix4UAV. Processing of the data was done three times using a different combination of GCPs each time to determine the impact of geometry and density of ground control on the accuracy of the DEM. Processing was carried out using 4, 6 and all 10 GCPs. The accuracies of the DEMs were checked using the GCPs measured with the RTK-GPS rover.

    Figure 6: Height difference between GCPs and DEMs generated from UAV point cloud data.

    In addition to assessing the internal accuracy of the DEMs generated using the collected UAV data, the DEMs were compared to an independent ATV RTK-GPS survey conducted by WRL from UNSW. A 50m by 50m area is shown to illustrate the difference in point density between the WRL survey (green circles) and the UAV survey (white dots). 

    Figure 7: Point cloud density of WRL (green) and UAV (white) data.

    Figure 8: Spatial coverage of WRL survey (green) and UAV survey (white).

    Three long sections were generated through the DEMs to compute height differences between datasets.

    Figure 9: Height differences between WRL and UAV DEMs from centre of beach long section.

    Figure 10: Height differences between WRL and UAV DEMs from lower beach long section.

    Figure 11: Height differences between WRL and UAV DEMs from upper beach long section.

    The ATV cannot drive on difficult beach terrain and therefore these areas are not measured using RTK-GPS. The UAV does not have this problem and is capable of collecting data for these areas. In order to assess how well the UAV can identify this difficult terrain, an additional experiment was conducted using RTK-GPS on foot.

    Figure 12: DEM profiles for the East-West cross section of the depression

    Figure 13: DEM profiles for the North-South cross section of the depression.

    The measurement precision of the UAV is severely affected by the ocean waves. The feature matching software cannot identify matching features from non-stationary objects (i.e. ocean waves).

    Figure 14: UAV DEM showing large variations in height at the shoreline.


    Results from this thesis demonstrate that further investigation into the quality of data obtained from UASs is needed to accurately assess its usefulness for coastal applications. Particular focus should be on identification of beach scarps and large depressions in the beach surface over short distances. The results of this thesis have been obtained from a single UAV flight and hence cannot accurately represent the capabilities of the UAV. This thesis does, however, show that there is potential for a UAV to be used for accurate 3D mapping of beach terrain.  

    copyright Jonathan Clancy 2013