Abstract:
The use of Unmanned Aerial Vehicles (UAVs) for traffic observation has increased with
advances in both drone technologies and video analytics. Improvements in video analytics
for identifying and tracking vehicles in two dimensions means that it is now possible to
collect and analyse vehicle paths and traffic movement dynamics. This is useful for
understanding vehicle movements through roundabouts. This paper will report on a large
study using UAVs to collect data at over 100 roundabouts across four regions in South
Africa, specifically considering the process of data collection using drones and the
software requirements for data extraction. Vehicle operations at roundabouts were
recorded by UAV-mounted cameras during peak traffic hours from optimal vantage points.
The footage was then uploaded to a cloud-based video analytics platform for processing
and scaled to the relevant coordinate system. The video analytics platform was found to
provide accurate measurements and movement dynamics of vehicles as they travel
through the camera view. Analytics that are directly available from the platform include
point speeds at specific locations, acceleration, deceleration, and gap acceptance. Traffic
counts and turning movements are also identified because every vehicle traversing the
roundabout is tracked. This paper will further report on the post-analysis techniques that
were followed to analyse the data and extract the relevant performance measures required
for further research.
