We take the watchful eye of surveillance cameras for granted as we go about our daily business in public spaces and even our homes. But the hours of CCTV footage collected can be an issue for security staff trying to identify suspects quickly. Dr James Orwell is leading the privacy-sensitive video surveillance project at Kingston University’s Digital Imaging Research Centre. They have designed a CCTV system to analyse footage automatically and speed up the process of finding security incidents. This project began when the team were contacted to work on a system that could be deployed in Milan Airport and on Spanish Railway.
Credit: The British Council
REAL TIME AND FORENSIC MODE
The science is based on pattern recognition technology and statistics, in two stages. First, is the training phase in which the system becomes familiar with the behaviour observed by the camera network. Once achieved, the second stage is to track a particular person’s movement over time. Dr Orwell points out that they can’t use face recognition because the images are too small: ‘we generally try to use the colour and composition of their clothing. So you form a representation of what they look like and you look for that in other images from other cameras.’
The system is designed to work in a real time mode, monitoring places where the system expects to see a person; and a forensic mode where previous footage can be examined for evidence.
There are plans to include an alarm bell mode. Dr Orwell gives the example in an airport of a person walking through a door the wrong way, or trespass in a secure area. He explains, ‘the system would then be used in a forensic capability. What we are able to provide is much faster in terms of locating the relevant images and a clearer image of that person.’
Credit: The British Council
CHALLENGES
Their project has faced tough tests, such as matching images of individuals with variations in camera angles and lighting, and monitoring individuals where shops and other distractions can lead to unpredictable gaps in footage. The biggest challenge, Orwell points out, is locating individuals in a crowd. In close partnership with Ipsotek Ltd, the team are seeing their ideas turned into an intelligent CCTV system that can tell the difference between a cat and a burglar.
Contacts and sources:
The British Council
REAL TIME AND FORENSIC MODE
The science is based on pattern recognition technology and statistics, in two stages. First, is the training phase in which the system becomes familiar with the behaviour observed by the camera network. Once achieved, the second stage is to track a particular person’s movement over time. Dr Orwell points out that they can’t use face recognition because the images are too small: ‘we generally try to use the colour and composition of their clothing. So you form a representation of what they look like and you look for that in other images from other cameras.’
The system is designed to work in a real time mode, monitoring places where the system expects to see a person; and a forensic mode where previous footage can be examined for evidence.
There are plans to include an alarm bell mode. Dr Orwell gives the example in an airport of a person walking through a door the wrong way, or trespass in a secure area. He explains, ‘the system would then be used in a forensic capability. What we are able to provide is much faster in terms of locating the relevant images and a clearer image of that person.’
Credit: The British Council
CHALLENGES
Their project has faced tough tests, such as matching images of individuals with variations in camera angles and lighting, and monitoring individuals where shops and other distractions can lead to unpredictable gaps in footage. The biggest challenge, Orwell points out, is locating individuals in a crowd. In close partnership with Ipsotek Ltd, the team are seeing their ideas turned into an intelligent CCTV system that can tell the difference between a cat and a burglar.
The British Council
0 comments:
Post a Comment