Video analytics can provide enhanced metrics on the level of activity of pedestrians and vehicles on our city streets.
This project builds on our existing CCTV work, which uses open-sourced object detection models to identify pedestrian and vehicle activity in images sampled from designated home positions. Working with the Urban Observatory at Newcastle University and Glasgow City Council’s (GCC) CCTV team, we will use the same approach to develop a framework to provide more accurate measures of activity and use of city streets by pedestrians, vehicles, and cyclists.
Video analytics uses object detection models, which are like those used in our CCTV project. When analysing pedestrian movements, these models identify people in a frame and then where they are in subsequent frames. This way, people’s movement through the image frame can be tracked, allowing for counts of individuals in different directions and traces of movement in the space to be measured.
Aims and objectives
The project aims to identify the most appropriate generalisable model for use across different cameras in Glasgow’s CCTV network.
The project team will:
- Explore the delivery of analytics in three distinct areas:
- Enhanced detection of pedestrians, counting moving and non-moving objects, identifying direction of movement.
- Enhanced detection of vehicles, counting moving and non-moving objects, identifying direction of movement.
- Development of street activity metrics, including tracing objects (pedestrians and vehicles) through space.
- Develop methods and software for implementation within the CCTV suite to operate on stand-alone machine(s).
- Assess the impact of the system on the Community Safety operation to identify and resolve any issues.
Lead: Dr Mark Livingston
Team: Dr Andrew McHugh, Dr David McArthur, Luis Serra, Maralbek Zeinullin
GCPH Team: Bruce Whyte
GCC Team: Kimberley Hose, Keith Scott, Kalim Uddin