Have you ever wondered how many times that you have appeared on a Closed Circuit Television (CCTV) system somewhere or other even if, for example, this might have been taken from a short trip into say Westminster on the London Underground? The number of times that you have probably been caught on CCTV may well be in the thousands since these first began to appear on our streets and in our public and private buildings thirty years ago but have you ever thought about how the analysis of what is captured on CCTV footage is done and how long this takes? A British company called SeeQuestor has been doing just that and it has come up with a way that makes the process of analysing CCTV footage very much easier, more efficient and cheaper.
Estimates vary, but well over one million CCTV cameras are thought to be in operation in Central London alone these days and there are estimated to be over 200 million CCTV cameras in operation worldwide. CCTV has to an extent changed the world and how we see it but in truth it is also a system that has a huge flaw because of the time that it takes to analyse the data footage itself by specialist people and who may be looking for specific evidence of crime that will assist them in the investigation. A completely separate issue that has long needed to be resolved is how the information gathered from the hundreds if not thousands of hours of CCTV footage collected can be best shared amongst those that require specific information?
Whilst it might be nice to imagine that CCTV footage is all electronically filtered and processed in, shall we call this a ‘Cross-Government Intelligence Centre’ and that, just as so often appears in various American and British TV detective series, this can be done in seconds on touchscreens held in a central location, the reality of how CCTV footage is analysed is rather different to what we so often see on our TV screens and far removed from ‘Hollywood’ style touchscreen concepts that appear to show analysis achieved in seconds.
The reality in terms of police detective work in the US, the UK and right across the rest of Europe is that most CCTV analysis is actually done by one policeman or detective trawling through hours and hours of video footage hoping to find answers or clues that will help solve a specific crime. Using the most simple of maths, when it comes to searching a user will need at least one man-hour of detective work for every one hour of CCTV video looked at. Put into context of say the London riots in 2011 which provided 200,000 hours of actual CCTV footage, one can begin to appreciate the labour requirements of such a mammoth and laborious task.
Thankfully, video intelligence and analysis technology is catching up and a new and very exciting forms of CCTV processing solution has arrived courtesy of SeeQuestor, a British start-up technology company based in small offices in the ‘City of London’. Recognising a need for faster and more specific analysis of CCTV data would provide in the security and policing activities SeeQuestor has designed an innovative collection of solutions to improve the efficiency of CCTV analytics.
Even if you don’t have a great depth of knowledge of technology or surveillance, the first challenge of CCTV analysis should be fairly obvious – the format. It may be hard to imagine in this day and age but CCTV systems in operation today use a bewildering array of different video formats, including hundreds of different digitally-coded files and even traditional tape formats that then need to be processed.
The SeeQuestor solution is to convert every disparate format to the industry standard – MPEG4 – which also happens to be the usual format you might already use on your laptop or smartphone. Conversion of CCTV footage is then achieved through a simple web-based platform or, on the more challenging format conversions, by using their own in-house laboratory technicians.
Whilst the ‘standardisation challenge’ has been previously addressed by some industry bodies and organisations, it is very surprising that as yet there is little evidence of a single organisation implementing or mandating a common video format. Nevertheless, there can be little doubt that the SeeQuestor solution really is an excellent answer to the long standing issue of format problems at least and until a universal recording standard can be agreed upon by all CCTV users.
Once CCTV footage is all in MPEG4 format and viewable, the issue of labour to process all information and search for the relevant events can then be addressed. That is where technology and computing power becomes a ‘force multiplier‘ of the human effort and analysis. SeeQuestor systems use filters for motion detection, person detection and face recognition, the technology for which is fairly established in the consumer market. Indeed, facial recognition has been used on smartphones and social media for many years but in the case of the SeeQuestor system it is clear that the operator interface provided is very much more advanced. For instance, it is possible specify a person in a certain colour of jacket, with a certain type of hairstyle or perhaps the person you are searching the CCTV footage for, may be known to have been wearing glasses. This can then be presented in what can be called a ‘virtual police line-up’ of all the various CCTV occurrences found on selected footage and importantly, all this being done in minutes rather than hours or days.
This new and very innovative form of computer based output can then if required have effective secondary analysis applied to it. For example, the extracts from the video can be assigned to specific cases (these are termed as ‘Quests‘) which allow for effective collaborative working on a specific project. Exact geographical locations of the CCTV cameras can also be plotted on a digital map allowing for timed reconstructions of the specific event and, if required, plotting the furthest that a ‘suspect’ could have travelled from a scene either on foot or by vehicle.
Combining all the various techniques offered by the SeeQuestor software system certainly makes this a very powerful analytical tool. Having seen this at first hand and heard from those that have trialled it, it is very impressive. Whilst a simple criminal incident could very effectively be processed on the system my understanding is that the experiences of a ‘pre-launch‘ customers in the UK, US and Australia have shown the potential for much wider applications being used. For example, though the exact details remain classified, the Arizona Police financial crimes task force have used the software in an Intellectual Property case valued at over £32m. Working from an untenable situation that required potential manual viewing of over 4,000 hours of footage with a small team of detectives, using the SeeQuestor system enabled them to achieve reliable video evidence in just four days.
Ethics & Responsibility are clearly very important aspects of examination of CCTV footage and none of this has been lost on SeeQuestor. Safeguards are built in and, as laid down in the UN International Covenant on Civil and Political Rights, the legal regulations on the collection and retention on CCTV data are very fully applied. Having appointed a ‘Chief Ethics Officer’ and in taking the whole issue of ethics very seriously SeeQuestor has understandably been keen to point out that the software is only used for post-event analysis as opposed to real-time monitoring/filtering.
Whilst users of the SeeQuestor software may eventually be able to pick out vehicles and further down the line, maybe even weapons too, the system has so far been designed primarily to identify humans through analysis of their various features. Clearly the potential for extension of the software to cover many other aspects of security, policing and detection work including extra searches is entirely possible and in the longer term, military applications could become possible.
Having implemented their own set of ethical policies the company accepts that these are more than sufficient to prevent export sales to states or organisations where civil liberties are likely to be compromised.
One of the things that I found most interesting is not just that this is UK based innovation but that it falls very much into the category of Science, Technology and Maths (STEM) products. We have come a long way in a relatively short period of time and the computing power required for the processing CCTV data and that is provided to customers to operate is the equivalent of the IBM Blue Gene supercomputer which was, in 2004, the world’s most powerful computer.
But is what SeeQuestor have achieved really to be considered as a technological advance? The answer is undoubtedly yes. And are there potential ‘trickle-down’ benefits for research and education here that could potentially boost STEM in the UK? SeeQuestor certainly believes there are. Indeed, though some of the technology may not be completely revolutionary, the company are committed to research programmes with Queen Mary University in London and the University of Cambridge as part of a ‘Deep Insight’ project to improve learning computers. I understand that this includes research using the NVIDIA GDX-1 Deep Learning Artificial Intelligence Supercomputer.
It seems that SeeQuestor may have picked a good gap in the market in video intelligence.
Has what SeeQuestor has designed got the potential to be a great success in a world of increased security requirements and need for faster crime detection? Undoubtedly so. Indeed, based on the experiences of the pre-launch customers and the potential of the system, the sooner it comes to market the better. Having visited the company and also, with my assistant Matthew Winwood, attended the formal launch which took place last month and having listened to some of those that have used this brilliant new software technology I would hope that rather than allow individual police forces in the UK to investigate the benefits and cost savings this brings, that it should be the Home Office that embraces and takes positive action to role this technology out across all police and security forces.
Whilst there are clearly some ethical concerns with the technology that remain, SeeQuestor is aware of these and will work with all the authorities involved. Primary amongst these are the often politically-charged questions of surveillance and civil rights. My hope though is that from a cost and efficiency basis alone this excellent fast intelligence based video analytics technology can secure use by all by all of the overstretched police and security forces in the UK without delay and that SeeQuestor and Britain can also reap large export rewards as well.
Howard Wheeldon and Matthew Winwood – London – 21st October 2016