Police could spend less time on paperwork and more time on the beat, if a UK project to develop a system for automatically extracting evidence from CCTV footage is successful.
Researchers from Kingston and Surrey Universities are working with the Police Scientific Development Branch (PSDB), the Home Office, and sensor specialist Sira to develop the surveillance system. It will generate a gallery of suspects’ pictures and vehicle number plates for all moving objects in video footage taken from CCTV cameras, said Dr Graeme Jones, director of Kingston University’s digital imaging research centre.
The team hopes to produce a prototype system by spring 2006, which will be capable of finding the clearest image available of each person and vehicle in the footage and tracking their movements, he said.
‘The machine will take in video tapes and generate visual information, such as the types of objects the principle entry and exit points of each person, and their speeds. Once they are marked up we can say a person entered a shop at door number one at this particular time. We can build a description of all the visual information in the scene,’ he said.
The project, called Reveal (Recovering Evidence from Video by Fusing Video Evidence Thesaurus and Video Meta-data) aims to computerise evidence gathering, freeing police time. ‘This is part of the bureaucracy of crime prevention, it is very labour intensive, and officers see it as preventing them from getting out and solving crimes,’ said Jones.
Researchers at Surrey University are working with the Met’s training centre at Hendon to gain an understanding of how forensic officers spot suspects or objects, said Prof Khurshid Ahmad, head of Surrey’s computing department. ‘I will be trying to determine how a person distinguishes between a suspect and non-suspect.’
Ahmad’s team will use transcripts of experts commenting on CCTV footage to build a thesaurus of terminology used by police.