The UK is one of the most surveilled nations in the world. An estimated 5.9 million CCTV cameras keep watch over our every move, but the volume of footage creates a problem: when the police or security services need to actually analyse it, things move very slowly. Despite the proliferation of CCTV, technology for handling the petabytes of data collected is still little more than human eyeballs and lots of patience.
That might be about to change. New technology could allow police and security services to quickly analyse CCTV footage to look for movement, faces and track suspects across the world. By linking 'dumb' CCTV cameras to a 'smart' online system, authorities will soon be able to find and track anyone. Trials of the technology with two UK police forces could begin in October.
In the aftermath of the London riots in August 2011 police scoured through more than 200,000 hours of CCTV to identify suspects. Around 5,000 offenders were found by trawling through the footage, after a process that took more than five months. Finding missing people is similarly arduous work -- when teenager Alice Gross went missing in September last year 30 officers were tasked with combing through CCTV from 30 cameras, covering a six-mile radius.
CCTV analysis mostly relies on teams of specially trained officers watching thousands of hours of footage, waiting for that one crucial second of evidence. It is a muddle of more than a thousand video formats, poor quality footage and manual processing. The system is at breaking point, but rather than investing in new technology, police forces have simply trained more officers to eyeball footage.
Much public debate has centred on the sheer number of CCTV cameras keeping watch on the UK. By one estimate people in urban areas of the UK are likely to be captured by about 30 surveillance camera systems every day. That's systems, not individual cameras. Unsurprisingly, police are struggling to handle the sheer volume of footage available to them when they start to investigate a murder or try and locate a missing person.
In its report into policing of the London riots of August 2011 the Metropolitan Police identified updating CCTV technology as a high priority. It said the investigation had "highlighted the extent and value of CCTV evidence" but also demonstrated the need "to do more to harness the value of this tactic". Slowly, the UK's police forces are realising they need to modernise.
In March of this year the government finally recognised something needed to be done. The Home Office established the Video Analytics for Law Enforcement (VALE) initiative in an attempt to try and find new technologies to replace the archaic tools on which police currently rely.
CCTV analysis is 'a disaster'
William Addison, director of video analysis SeeQuestor, describes the present situation as "a disaster". His company is currently developing a system that it claims can make any CCTV camera in the world smart. "At the moment it's one hour of eyeballing for one hour of footage," he tells WIRED.
Current video analysis techniques are poor, relying on expensive, specially-installed cameras to make things like facial recognition and person tracking possible. The likes of OmniPerception and Ipsotek have sold CCTV analysis systems for years, but SeeQuestor claims its technology is different as it works with any footage from any camera, rather than relying on specialist hardware. "The challenge is the source of the footage. If a crime happens in London, or wherever, there's absolutely no guarantee for the quality of the footage and the types of cameras that are being used to generate that footage. Whether those cameras are calibrated is also highly unlikely. And it isn't just about where the crime happens -- you want to track the person, find out what happened, who else was involved," he says.
SeeQuestor's system, which will be ready for police to trial from October, takes 'dumb' video and makes it smart. The company has worked on three analysis tools for the first trials, but says others could be developed. "We're engaged very heavily with the Met, with Scotland Yard, with counter terrorism in the UK, with British Transport Police -- these are the largest users of CCTV," says Addison.
WIRED contacted the Metropolitan Police and British Transport Police about their potential use of SeeQuestor's technology or any other CCTV analysis tools but neither force was willing to comment for this article.
Tristram Riley-Smith, founder of SeeQuestor, confirmed the company was "involved in negotiations with two police forces," in the UK with a view to starting trials in October. Riley-Smith, who was a senior liaison officer at the British Embassy in Washington DC after 9/11, described current CCTV analysis systems as "incredibly archaic". "You'd be amazed. It is incredibly cluttered and very, very slow".
Finding a needle in a haystack
It starts with a simple task -- converting any footage into MPEG4. Even this basic service could make a big difference -- police currently struggle with more than 1,000 video formats from the millions of CCTV cameras across the UK.
Once converted, the footage can be analysed. SeeQuestor's software can pick out faces in crowds, detect movement and track suspects or missing persons across the world. All it needs is the basic footage from any 'dumb' camera. The company claims its system can analyse footage for faces and movement up to 12 times quicker than a human. "What analysis can you usefully run on footage which is generated from any camera? That's the challenge," Addison explains.
WIRED is taken through a live demonstration of the technology using CCTV footage captured during the 2011 London riots. In the footage we see a backdoor on a quiet street, from which it is hoped someone will emerge. Having analysed the footage the software creates a timeline with every second of motion it has detected before splitting it into two categories: high probability (right in front of the camera) and low probability (further away in the field of view). An officer can then skip through all the sections where motion was detected.
In another piece of footage WIRED is shown people emerging from a Tube station in London. This time the system has been told to look for faces. On one side of the screen we see the footage and on the other we see thumbnails of every face detected. The faces are also grouped, so all instances of one person will be gathered together. Clicking on a still image of a face will skip the video to the exact second they were spotted.
The third tool is the one Addison is most excited about -- person tracking. Again WIRED is shown footage from the London riots, this time of a group of young men in an alleyway. A man wearing blue jeans, a white t-shirt and a black jacket rides into frame on a bicycle. This man is selected by the human operator, creating a coloured box around him.
The system then looks through all the footage it has been given to find instances of other people in blue jeans, white t-shirts and black jackets riding bikes. As with facial recognition it then displays likely matches in groups, this time from most likely to least likely. A human operator can then look through all the relevant of footage to see if there are any genuine matches.
"Computers are very good at crunching data and throwing out results of that data, however they're not so clever when it comes to identification," Addison says. "For that we do need the human eye to say 'right, this guy with the red t-shirt is not the same as this guy with the red t-shirt'."
Any changes or notes added to CCTV footage by the system are carefully audited and saved in a "black box", providing a trail that Addison claims will ensure evidence gathered stands up in court.
SeeQuestor's system relies on some fairly hefty technology and company plans to sell it complete with the necessary hardware to run video analysis. It estimates the entire setup will cost £200,000 for the software, hardware, five years of licensing and support. The company could also establish its own data centres to analyse footage on behalf of police and security services, according to Addison.
Where we go, who we meet, what we do
More cameras, new technology -- but the same regulations. Guidelines around CCTV in the UK were last updated three years ago as part of the Protection of Freedoms Act 2012. The regulations, which are not legally binding, only advise authorities on how they should behave. Not only is the act weak, it also risks being rendered useless by the "breakneck" speed of technological innovation, according to one surveillance expert. "This lack of debate and safeguards becomes all the more striking as technology overtakes even our most far-fetched predictions," Rachel Robinson, policy officer at human rights advocacy group Liberty, tells WIRED.
Despite the Snowden revelations about GCHQ and the NSA, the UK is yet to engage in a debate about the near six million CCTV cameras that watch our every move. Robinson warns technological advances in video analysis could "fundamentally alter the relationship between individual and state and have a chilling effect on freedom of speech and assembly." "When paired with new technologies -- movement detection and facial recognition in this case -- it has the potential to be a mind-blowingly intrusive tool, providing an incredibly detailed picture of where we go, who we meet, what we do."
Robinson argues SeeQuestor's technology would allow the authorities to join up disparate networks, "significantly extending surveillance capabilities and analysing huge amounts of information about the population at large". While the end result is more targeted, effective surveillance, the system still relies on analysing face and movement data en masse to track down suspects. Current regulations, she argues, fall worryingly short.
Earlier this year surveillance camera commissioner Tony Porter said it was important to ensure CCTV did not "proliferate unnecessarily". "Surveillance can be an extremely good thing and run well, it's a useful tool for society," he told the BBC in January. "You can still maintain the balance of excellent surveillance but not have a propagation of surveillance that is actually useless." "Where technology exists and comes to the fore it can support law enforcement and protect society. That's a good thing. But what my concern is about is the introduction of poor surveillance that doesn't add benefit to society."
Porter's comments were intended to spark a debate about the number of CCTV cameras, but technological developments could soon shift the focus. WIRED contacted Porter for this article but his office informed us he was on holiday until late August. Nobody else was able to speak on his behalf.
SeeQuestor's Addison claims his company's system has been designed to improve police efficiency and replace creaking technology. "We want to make police and security teams more productive so they solve more crimes and eventually do it with fewer people". But with public awareness of CCTV so poor, innovation risks putting people's privacy at risk.
For Robinson, CCTV is only useful when "necessary and targeted", with open, public debate "urgently required" in the face of new technology. "If we allow this chasm between technological capabilities and legislative regulation to continue growing, we risk opening floodgates that it will be extremely difficult to close."
This article was originally published by WIRED UK
