Researchers from the University of Wollongong have successfully demonstrated an algorithm that detects fights through CCTV cameras on public transport.
The team, based at the SMART Infrastructure Facility, has successfully demonstrated the ability for their software to detect a fight using actual footage.
Next, they will stage a “fight” at Wollongong station and test the ability of the CCTV network and operators to act on a violent incident.
Lead researcher Dr John Barthelemy said this was the first time that this type of artificial intelligence would be used by a public transport operator.
The project, which started in September, was one of four winners of a Safety After Dark challenge created by Transport for NSW.
The challenge focuses on making women feel, and be safer when travelling on public transport. Research into women’s safety revealed that girls and women do not always feel safe travelling in our city at night.
“The next stage will be like moving from the lab to the real world,” Dr Barthelemy said.
“We want to test how easy it is to deploy it in a station – we know that the core of it works but we want to test all the things around it.”
The software is designed to automatically analyse real-time camera feeds and alert an operator when it detects a suspicious incident or an unsafe environment.
The data and reports automatically generated by the software can then be used to help prevent the abuse and violence committed towards women after dark in public transportation.
The software uses an open source code to predict a fight by looking at typical human poses. A human controller who accepts or rejects the suggestion then reviews footage.
The result is then used by the software to analyse images with greater accuracy.
“I really want to keep working on the solution and how we can make an impact with it,” Dr Barthelemy said.