With the increased demand for the application of video analytics in both old and new CCTV systems, we set out to find some guidelines on how to future proof the investments both in terms of the transmission system and the analytics software.
There is much debate with regards to how efficient the use of CCTV is, not least in terms of crime prevention. This to a large extend is down to the fact that CCTV footage often has been of poor quality. However, it is now predicted that the use of video analytics software will significantly change the perception of CCTV. According to Graeme Gerrard, Deputy Chief Constable of Cheshire Police; "The contribution of CCTV to the detection of crime is likely to equal that of DNA and fingerprints."
The increased use of video analytics can safely be deemed the 'next big move' in CCTV and the market growth for this emerging technology bears testament. The figures are documented by the latest research from Frost & Sullivan where Dilip Sarangan, Industry Analyst for Physical Security Group explained, "Video Analytics is a relatively new technology and our research only started in 2004. However, we expect the World market to grow from US $145.7 Mill in 2008 to US$ 416.3 Mill in 2012. That equals an annual compound growth rate from 2005 to 2012 of 31.9%. If we look at the equivalent figures for Europe, Middle East and Africa they show an annual compound growth from 2005 to 2012 of 29.9%."
As with any emerging technology there is much confusion; when is the right time to invest, where should the video analytics be located, how do you gain the most from investing in video analytics applications - are just some of the issues the end users as faced with.
We have spoken to Chris Gomersall, CEO of Ipsotek, a UK based company specialised in the development of Intelligent Video Analytics Software and Dr. Alan Hayes, Founder and MD of AMG Systems, a manufacturer of CCTV transmission systems, to get some guidelines with regards to what users should be looking for when considering the implementation of video analytics on existing or new CCTV systems.
Image quality: alpha and omega for scene analysis software
Chris Gomersall said "Let's start by looking at where to apply the software. Basically there are 2 options: At the camera point or in the control room. Add to this the consideration that good analysis requires maximum video content."
"Video analytics can help to enhance the CCTV operator's effectiveness by alerting them to events, live as they happen, allowing them to take action quickly to intervene or call the authorities. However, it is vital for the effectiveness of any software solution that the quality of the image is useful for the purpose", Chris Gomersall explained. "It is worth pointing out that the 'quality of image' issue not only applies to existing CCTV installations, where an upgrade with video analytics software can increase the efficiency in terms of using CCTV images for forensic evidence to help solve crimes. Likewise, for new CCTV projects there is a requirement to ensure that future proofing takes high priority. This becomes even more important because - as has already been pointed out - the push towards IP based CCTV systems in itself, creates a limitation because of the compression which is introduced to the video signal."
Compression limits the usability of CCTV Systems
"It is evident that for systems where pre-event analysis takes place, either using traditional security operators in a control room or the more recent pixel based analysis software, the reduction in video quality caused by compression can severely limit the usability of the CCTV system", said Chris Gomersall.
Dr. Hayes supplemented, "Video compression unfortunately compromises what you can do. In a highways system for example it affects traffic management capabilities and others such as security, which is where the market is heading right now. When you compress CCTV images for transmission, you're essentially throwing away low value information. If nothing changes from frame to frame, then a compression-based system will see no reason to re-send an image. If you get motion then you send the changes. Hence with compressed video you get basic motion detection. What you don't get is pixel-based recognition. This is why Ethernet based solutions need analytics at the camera, not in the control room."
Chris Gomersall: "The motion detection approach is fine if you can do what you need to when you buy a system. But it means that upgrades are a problem, because as processing power increases, eventually you will have to upgrade on a per camera basis - and that could potentially become a very costly affair".
Consider the security aspect
"Another important consideration from the security aspect is that Ethernet systems can be hacked. It is possible to break into a system at any point and inject your own feeds - there are actually documents on the internet describing how to do this. Of course you can put firewalls in place but these can also be penetrated", said Dr.Hayes. "It is also worth considering that Ethernet systems require skilled maintenance distributed to all the remote locations. If the system breaks down, it can take a very long time to get the network up and running again. Further, migrating to next generation you could potentially be looking at a complete swap of remote equipment."
"For the transmission system there is good news. Even for large wide area networks it is now possible to use uncompressed digital video transmission, such as recently implemented by the Museum of Islamic Art in Doha in Qatar. This transmission technology replicates the capabilities of an Ethernet system and transmits the entire content of the video images back to the control room. It also allows for drop-and-insert of camera signals - as and when required - as well as dual redundancy. It even carries Ethernet which for example could be used for simultaneous retrieval of DVR images", said Dr. Hayes.
"There is a significant point to be made with regards to what you want the analytics to detect for you", said Chris Gomersall. "This is why we prefer to talk about intelligent video analytic. If for example you want to track a drug deal in a nightclub – where it is noisy, dark, and full of people behaving slightly silly, you can actually program the analytics for each individual camera to track certain behaviour. One camera might be equipped with software capable of tracking behaviour associated with handing over drugs, whilst other cameras might be equipped with software tracking for other things such as aggression, facial recognition of wanted criminals etc. Compare this to an application which could, to some extent, be considered exceedingly boring for an operator – watching the London Eye 24/7 in order to make sure no-one attempts to climb it. Now, there is obviously motion in the picture already as the Eye is moving, so you need an intelligent solution capable of recognizing a human climbing the wheel which will then raise an alarm in order for the operator to take appropriate measures."
Dr. Hayes concluded, "For 'observation critical' infrastructure and security applications, AMG always recommends that the CCTV images are brought back to the control room in uncompressed format. This approach future proofs the CCTV solution and caters for the application of Intelligent Video Analytics."
The Video Analytics Check List
1. If you are looking to increase the level of surveillance without increasing manpower Video Analytics is a cost effective solution as it allows you to install more CCTV cameras and monitor them without increasing manpower.
2. Video analytics works best on uncompressed video in the control room.
3. Video analytics can often retrofit into existing systems, using existing cameras. Hence it improves the ROI and efficiency of existing CCTV installations.
4. Video analytics is good for camera views where there is little regular activity. It allows an activity to be flagged up for the attention of the operator and thereby reduces the risk of overlooking a "dull" camera point.
5. Intelligent video analytics systems are individually programmable on a per camera basis. This reduces the number of false alarms as the built in intelligence know how to deal with a busy scene.
6. If you have a system which requires monitoring of several events simultaneously, effective analytics systems have an ability to intelligently decide if an alarm should be triggered or not.