Is video-over-IP already obsolete?

The evolution of products within the CCTV market has been characterised by a slow but nonetheless steady progression. Recording systems have developed from the traditional analogue video tape-based solution to today’s high-tech digital video recorders, while cameras have moved from monochrome, tube-based acquisition to the latest DSP-based, backlight-compensated CCD modules.

Over the past few years, of course, there has also been a significant change from completely analogue-based solutions to hybrid set-ups wherein traditional analogue cameras interface with the new breed of digital video recorder.

In terms of spatial resolution, CCTV cameras have largely been based around the standard broadcast specifications (PAL in Europe and NTSC in the States). This affords a maximum spatial resolution of just under 800 x 600 pixels which, in most cases, is good enough to show that someone is committing a crime. However, it’s not enough resolution to make a positive identification in every case. That is only compounded by criminals’ current penchant for wearing hooded tops and masks.

Meantime, the requirement to transmit live video to either a recorder or video wall has necessitated fixed installations with very expensive set-up costs. The actual cost of the camera is often less than 5% of the total hardware and installation outlay required to receive the data at the video monitor or DVR. The need for fixed cables has also prevented the use of temporary cameras to monitor ‘hotspot’ areas which can come and go very quickly as police and security officer assets are deployed in response to the problem.

The evolutionary scale

Looking to the future, the completely digital solution of video-over-IP – combined with network video recording – is believed by many to be the ‘next big thing’ for end users. However, it would appear that video-over-IP could well prove an evolutionary cul-de-sac, with this particular market likely to be by-passed before it really begins to take hold.

The reason why I make that last statement is that all of these solutions – from the very first tube camera attached to a simple monitor through to the current all-digital video-over-IP system – happen to share the exact same fundamental paradigm: video information has to be transported from the camera to a recorder and/or a human systems operator.

Today’s video standards are a genuine throwback to what was achievable in the 1950s analogue TV world. There is no fundamental reason why a spatial resolution equivalent to the PAL specification – or a temporal resolution of a frame every 25th of a second – is the correct answer to a requirement. In fact, if you think about it, the very fact that CCTV cameras are fitted with expensive PTZ mechanisms ought to tell you that the variation in spatial resolution required is often much wider than can be achieved with a PAL camera and fixed focal length lens. All of this begs a question... ‘What is the most likely failure – the solid-state electronic camera or the PTZ system?’

Video is an extremely high data content medium, which is why video-over-IP solutions use compression in an attempt to manage the bandwidth. An uncompressed PAL resolution image is about 0.5 megabytes. As it is commonly accepted that video frame rates for real-time images need to be around the 25 or 30 frames per second (fps) mark, a single uncompressed PAL camera is transmitting around 12.5 megabytes of data every second.

That is difficult to achieve over an Ethernet network... When one multiplies the amount of data by the number of cameras needed for a security system, then the bandwidth requirements – assuming that one wanted to retain uncompressed video – become unmanageable (to say the least).

Taking a step backwards?

In terms of the technology, surely we are stepping back to try to overcome this problem by using cameras with lower resolution than PAL (be it CIF or 4CIF resolution), often with lower frame rates and with the additional image degradation caused by whatever compression method is being used?

What we should be doing is finding the correct solution to solve the problem – whether this is CIF resolution or using 10 million pixels. Two frames per second might be adequate in many scenarios, but for others (such as spotting the knife being used in an altercation) 100 fps might be required to ensure a full understanding of what actually happened.

The time has come for all professional security managers to question that basic paradigm which, after all, gives us the phrase ‘video-over-IP’ – why do we want to transmit ALL of the video ALL of the time? More than 99.9% of CCTV video frames have no interesting data within them. No banks are being robbed. No OAPs are being mugged. In truth, we only transmit the frame because we do not know if it has relevant information within it or not.

The time has come for all professional security managers to question that basic paradigm which, after all, gives us the phrase ‘video-over-IP’ – why do we want to transmit ALL of the video ALL of the time? More than 99.9% of CCTV video frames have no interesting data within them

There are already intelligent systems on the market that undertake video content analysis, developed by companies including Activeye, Cernium and ObjectVideo. In many ways these systems are now at the same stage of development as the early DVRs. Essentially, they are high-end PCs running image processing algorithms which search out ‘suspicious behaviours’ and highlight them to a human operator. They still conform to the problem paradigm in that they have to have the cameras feeding them image streams, but they are capable of deciding what might be relevant video while in turn ignoring the mundane.

‘Smart’ cameras: an overview

If we look outside the world of CCTV to other industries that make use of automatic image processing, the machine vision market for industrial automation is already using so-called ‘smart’ cameras in significant quantities.

‘Smart’ cameras contain a built-in microprocessor whereby the image processing or video content analysis is actually done on board and only a pass/fail (or similar output) is generated by the camera. Many of these cameras incorporate secure digital memory slots allowing Gigabytes of storage – at low cost – to be implemented within the camera which then becomes its own DVR. The ‘smart’ camera can therefore store the last/next x minutes of data to allow both pre- and post-event incident review if necessary.

A similar system approach is now possible with security video, with only potentially relevant information being passed along for off-camera recording and human review. This would mean that in a 40-camera environment perhaps a maximum of only one or two cameras would be transmitting at any given moment in time. This would reduce the data bandwidth and storage requirements, or alternatively allow each camera to transmit a good deal more information – either in megapixel resolution, uncompressed video or perhaps 100+ fps (depending on the specific application, of course).

This technology is already here. Currently, the ‘hottest’ market for video content analysis is Automatic Number Plate Recognition (ANPR). The Spike camera developed by PIPS Technology and the Vega unit are true ‘smart’ CCTV cameras where the ANPR algorithm is run within the camera. This offers the end user a number of potential advantages over PC-based ANPR systems.

For one, the total power requirement drops significantly (in some cases to just 7 W). That opens up the possibility of harnessing solar or wind power sources. Also, on the basis that you are actually using the ANPR system to recognise number plates, the amount of data that has to be transmitted is very small. Just a short text string. This can be achieved by using wireless or even mobile telephone infrastructure, from Bluetooth through Wi-Fi to microwave mediums.

In short, it is now possible to deploy a real-time CCTV monitoring system extremely quickly at almost no cost. This technology will open up a whole range of applications that, in days gone by, have never been cost-justifiable for the average security manager.

IP: is this just a sideshow?

ANPR will be but the first of a number of applications. The others will migrate from the PC systems world, and will then be ported to ‘smart’ CCTV cameras. Such cameras also offer the advantage of removing the ‘single point of failure’ problem. If one camera fails it doesn’t affect the rest of the system.

Given that the market for video-over-IP has yet to take off, I would argue that there are still significantly greater numbers of installations using analogue cameras than those using IP-based systems. There already exists a replacement technology that overcomes the major technical deficiency of CCTV, and which is more cost-effective. It is also already starting to be deployed in real world applications.

One has to wonder whether the video-over-IP market is destined to become anything more than a security sideshow?