The Home Affairs Ministry wants those surveillance cameras put to more pro-active use, with new technology that can detect a face in a crowd, for example.

Last month, it made public two 'request for information' (RFI) documents asking about two similar types of technology: one that can recognise suspects by their facial features and another that can pick out objects of interest - from video footage.

In the document on facial recognition technology, it is looking for technology to help pick out suspects walking indoors or outdoors either by themselves or in a group, within view of the camera.

Essentially, it wants software which can compare images from the video footage to blacklist data of images from still photographs and or video clips. This is known as facial recognition technology, which records the spatial geometry of distinguishing features on a face.

In the second document, it is looking for technology that can help search for a person, vehicle or object from archived closed-circuit television (CCTV) recording. 'Such post-investigations usually require CCTV operators to review many hours of CCTV recordings,' it said.

Having this technology would shorten investigation time as investigators need only extract relevant parts of the recordings for analysis.

Dr Terence Sim, a researcher on facial recognition technology, likened the two technologies to Internet search engines.

'Instead of typing in the text of interest, the user supplies a face image or clicks on the object of interest in one part of the video, and the system searches the rest of the CCTV video for the person or object,' he said.

He noted that British police were able to track those behind the London bombings three years ago because of surveillance technology.

Passers-by handed them cellphone pictures of the suspects' faces. Also, all London subway stations had CCTV cameras.

They compared the phone pictures with the surveillance images, a process that could have been faster if facial-recognition technology then was good enough and accurate enough to search the videos automatically, Dr Sim added.

First introduced in 2003, there are now more than 120 CCTV cameras keeping a round-the-clock watch on places like Boat Quay, Orchard Road, Shenton Way, Little India and Geylang. These 'live' images are then sent to a command monitoring centre that is manned 24 hours.

More are expected to be added, given that police had said at the start of this year that they were looking to add a further 223 cameras to the streets.

The cameras have been instrumental in solving some crimes.

For example, police officers scouring through video footage managed to crack a 2004 case at Newton Food Centre, where a group of 10 men armed with parangs, samurai swords, knives and metal rods attacked two men.

Terrorism expert John Harrison saw the ministry's bid in seeking the new technologies as the 'logical next step'.

The presence of CCTV cameras may not deter determined criminals, he said.

'But if the system has an ability to detect a threat, item or individual, it increases the deterrence and perhaps prevention potential (of crime),' Dr Harrison, an associate professor from the S. Rajaratnam School of International Studies, added.

Dr Sim, an assistant professor at the National University of Singapore's school of computing, held a similar view but suggested that as the technology matures, hard core criminals would become more sophisticated to avoid being spotted.

Such technologies are already being used here, though not for crime-fighting purposes.

For example, local-based facial recognition technology firm XID Technologies has deployed facial recognition systems at several foreign workers' dormitories, controlling the access of these workers.

Abroad, facial recognition technology was installed in Tampa, Florida in 2001 so that the authorities can weed out potential terrorists among the crowd at the Super Bowl.

It was also used in the district of Newham in England to fight crime. In fact, the crime rate fell after police installed several hundred surveillance cameras connected to facial recognition technology.

There are, however, limitations to the technology available today.

For example, Dr Sim said that comparing videos from two different sources can be difficult, given the differences in illumination, camera hardware, viewing angle and so on.

He added that in Florida's case, the system was not accurate enough, detecting many 'criminals', who weren't. Till today, the error rates for surveillance face recognition are still high, he said.

Then there is also the issue of privacy, which Florida and Newham had to deal with when these places deployed the technology.

Dr Sim, who is also the chairman of the Workgroup on Cross-Jurisdictional and Societal Aspects of Biometrics in the Biometrics Technical Committee, said privacy issues needed to be considered.

Some questions he posed: Should covert monitoring of an individual in a public space be allowed? Should monitored public space be clearly marked out so that passers-by can be alerted? If they are alerted, does it mean they consent to being monitored?

'I think we need answers to these questions before facial technology becomes sufficiently mature and widespread. Otherwise, we will encounter abuses and even oppression on some segments of society,' he added.

carolynq@sph.com.sg

Read also: Rise of security cameras in recent years