Analysis of CCTV footage through use of algorithms to characterise proximity and surface contact on different transport modes.
Led by Newcastle University
The mechanisms through which SARS-CoV-2 spreads mean that transmission depends on the distance between people, their contact with different surfaces, and the number of people in confined spaces. This work package aims to quantify those factors using data provided by transport operators such as ticket sales, GPS traces, and CCTV recordings.
We will establish:
- A workflow for processing CCTV data that respects privacy as much as possible, such as by obscuring timestamps and blurring faces.
A detailed estimate of the likely number of passengers in each vehicle at different times of the day, depending on overall patronage as different restrictions are in effect, and the distribution of those passengers within the vehicles.
A machine learning toolkit to automate analysis of the CCTV data with validation against manual processes. This will allow for batch processing of downloaded data and real-time processing in a small number of vehicles in the later stages of the project.
Probabilities for the distance between passengers at different journey phases including in stations and on platforms, and contact with different surfaces such as handrails and stop buttons.