Threatening behaviour on buses could soon be tracked and confronted by remote security analysts to protect the public from assaults and other criminal activity.
This is the hope of researchers at Queen’s University Belfast who are developing intelligent software to recognise anti-social behaviour from real-time CCTV feeds.
Working in collaboration with the Applied Criminology Centre at
The researchers claim the software will make a significant impact on preventing crime on public transport. Dr Paul Miller, head of the research programme, said that currently the UK’s four million CCTV systems fell short of doing this.
‘My personal view is that this type of smart technology has to be a differentiator in future security systems,’ he added. ‘There are concerns at the moment with the quality of cameras on buses… I wouldn’t invest in a CCTV system right now because I know it wouldn’t work.’
Miller said the industry was struggling to manage the large quantities of data that existing CCTV systems produced. As a result, recordings are often used to prosecute criminals following an event but are rarely able to prevent a crime before it happens.
The team at Queen’s believes its algorithms could tackle this problem. The algorithms work by analysing a passengers’ profile and combining their data with information about their movements, the bus’ location, the area’s crime statistics and the time of day.
Miller explained that the system would only be activated once the sum of the smaller events added up to establish a high-risk threat. Potential triggers could include loitering on stairwells, a group moving towards an individual and a passenger falling over on the bus.
‘The system will not necessarily say there is an incident occurring,’ said Miller. ‘But it will push that video to the top of the queue. The security analysts will then make a decision as to whether there is a situation and if they want to intervene.’
As well as on-board camera and screen technology, similar technology deployed at bus stops will capture an individual’s gender.
Their features will then be reduced in pixel size to produce a set of 50 coefficients. These act as a three-element vector through which the system draws a binary line to determine whether the individual is male or female.
So far, trials conducted on 4,000 faces have produced an 85 per cent accuracy rating in determining gender. According to Miller, when this data is combined with audio, the accuracy of the system increases to around 97 per cent.
Further work is being carried out on body shape and colour of clothing and the team is confident of deploying the system within the next five years.
‘This technology also ties into things such as the green problem,’ said Miller. ‘The government is very keen to get people out of cars and onto buses, but people won’t do this unless public transport is safe. Our system will facilitate secure transport corridors to encourage this change.’
The project is one of a range of research programmes being undertaken at the Centre for Secure Information Technology (CSIT), which was launched in September 2009. The centre intends to be at the forefront of commercial security systems development and has already received support from more than 20 industrial partners.
Homepage image: Copyright Transport for London 2005