Human Identification with CCTV

This article was originally published in the Sep/Oct 2004 edition of the Security Insider Magazine

The Challenge

The world of Closed Circuit Television is dominated by techno speak in relation to the electronic hardware components that comprise a CCTV system. A typical buyer is bombarded with impressive sounding technical specifications and jargon of claimed camera performance, DVR speeds and resolution and archive. There is much importance placed on these figures by manufacturers as they quickly point out how their camera or DVR is better than other brands, with the figures to support these claims.

There is no debate that these figures are of some use in assisting in the purchase decision. Like all advertising, the devil is in the detail, and as such these claims need to be understood for what they are.

I wish to suggest some other issues that should be understood by those who want to get the best bang for their buck in a CCTV system.

For many users the purpose of CCTV is to:

1. View (live) or Review (playback) events to understand what is or has happened. The extension of this is then to;
2. Identify persons involved so that action can be taken. This may vary from wishing to identify persons so they can be arrested and charged in relation to events, through to wishing to identify them so that lost items may be returned to them in, such as happens at baggage screening points at airports. There are of course many other variations of this theme, however it is fair to state that the majority involve the desire to identify a person.

While for most applications the first of these two requirements is met with existing CCTV systems, it is fair to say that the second point is a dismal failure. This is understandable as these two purposes are conflicting in their requirements for camera placement and lens choice. Live viewing or playback often involve looking at wide angle views in order to understand what has happened, that is reconstructing the event in the reviewers mind in an objective manner. For this often a wide view is the better choice. However, this means that a camera that provides this overview will never be able to provide the detail necessary for human identification.

By understanding that both of these objectives are to be met and designing a system to meet them, a CCTV system owner will reap tremendous benefits with their system for a margin increase in cost.

The key is to understand that there is a need to identify humans in almost all CCTV security systems, and to do this the system requires Target Specific camera placement and lens choice. The purpose of this Target Specific Camera is to do nothing else other than provide imagery of sufficient detail in order to allow the likelihood of Human Identification.

The Choke Point

In order to gain Human Identification with a CCTV system a very narrow field of view is mandatory. Contrary to what some camera manufacturers will say, no security camera will provide high enough resolution to identify humans if the field of view is too wide. What is required is for persons to pass through a narrow area covered by the Identification Camera in order to capture this required narrow field of view. We refer to these narrow areas as Choke Points. A Choke Point is often a doorway or narrow corridor or walk through metal detector thus ensuring we are able to frame the cameras field of view to the required minimum width.

Often the front entry door is the only Choke Point available as people enter a building before they fan out to different lift lobbies or other areas. A front door as a Choke Point often introduces significant back-light problems for the camera, with the high contrast as the persons passes from bright daylight outside to the internal lighting of the building. Where possible this situation should be avoided, as no matter what the manufacturer will say, strong contrast will always cause the image to be worse than the same camera placed with a choke point with more even light distribution.

The second issue to be careful with is the Transit Time for persons moving through the Choke Point. As we have discussed, in order to gain the Forensic Level identification a very narrow field of view is required. Accordingly it is a fact that a person can transit this area much more quickly than if the camera provided a wide vista. Thus the Choke Point camera must be configured on the recording infrastructure for a high record speed. If Video Motion Detection is used, as is often the case with DVR's, then the VMD must be set to ensure all persons transiting the Choke point are recorded. On lower performance DVR's this may involve using a separate PIR motion detector wired as an alarm input, rather than rely on the VMD, as some systems cycle time on the VMD is too slow. (The author has seen some DVR's where the trigger time for the VMD is over one second, which is much too slow for a person deliberately running through a Choke Point.)

Alternatively there are some very good DVR's that allow camera detected motion to be used as a trigger for associated cameras. Thus Camera 1 with a wide field of view in front of a building can be used to trigger motion based recording for Camera 2, which is the Forensic ID camera located in the Choke Point inside the front door way.

Typically for a Choke Point a record speed of 5 to 10 images per second is desired, at high to very high resolution.

How To Achieve This

In the typical CCTV installation the camera placement is high, with the camera fitted with a wide angle lens. ("so we can see everything that goes on") Sadly the opposite is true. By using a lens with a wide angle field of view detail is lost as the benefit of a wide vista is gained. Yes you can see that Person X came into the building then went to the Number 1 lift lobby, but the image of Person X is so small that it is hard to tell whether they are male or female let alone identify them or have images sufficiently good enough to narrow down the list of suspects.

The above images represent similar file size and quality, however the different result is a
factor of camera position and tightly defined field-of-view

What is required is a lower mounted camera with a very narrow field of view aimed at and focused on the Choke Point, ensuring that all persons entering, including those wearing caps, are captured in detail.

Mounting the camera low often means moving the camera back from the immediate area so as to achieve the front on horizontal framed close up field of view. In the Broadcast Television Industry this is called the Mid Close Up shot. (MCU)

Of course this is harder for CCTV installers to do, and the required use of a high range varifocal lens or mini zoom lens is more expensive. These two matters are often not part of the quote process or negotiation during the buying of the cctv system. The company that has sold the CCTV system wants to maximize its margin by using low cost wide angle lenses and have their staff or sub-contractor install the camera off the ceiling as this is the fastest way to get the camera up and operational. If wall mounting the camera at a specified height and using a 36mm auto iris lens is not quoted on than you can be certain that these two items do not happen!

Would you watch the nightly television news if the presenter was show from a camera mounted off the studio ceiling looking down at 45 degrees with such a wide view that you had trouble identifying them? Yet we expect our security staff to operate from a system that provides this view! And then think that if we go out a buy a more expensive DVR then this will be able to provide the imagery that we need.

The saying Garbage On Garbage Out is true.

The cameras field of view is the single most determining factor in respect to the usefulness of a CCTV system in identifying a human. Irrespective of the camera or DVR used.

A well placed ordinary quality camera with the correct field of view will always be better than the latest but poorly placed "super cam" with a wide angle lens.

The miniscule technical differences between high end cameras and quality DVR's are such that the choice often comes down to the buyers preferences and the local installation/support as determining factors in the purchase decision. What is the point of using such high end and expensive CCTV equipment when the camera is positioned so badly and fitted with a wide angle lens that no recorder in the world is capable of delivering the results that the buyer really needs. Even if you were to connect a real time broadcast studio recorder to these cameras the results would still be useless for human identification. What hope do you have with even a good CCTV DVR?

The British and European Standard BS… states that for human ID the field of view shall represent 75% of the height of a human. Thus the camera view shall be framed so that a person appears on the monitor from the top of their head to approximately their knees. This simple technique ensures that the required identification is achieved. Interestingly the proposed Australian Standard is following a similar path, thou it is likely to be for 100% of the target person. That is that when viewed on a monitor the persons fills the image from the top of their head to their feet.

The Result

For a typical office or industrial building all entry points into the building should have a camera dedicated to this MCU view. That is a camera for every door in or alternatively at a better internal choke point where back light will be less of a problem and where the one choke point serves many entry doors. (A metal detector screening point is a good example of this)

Should an event of interest occur, be it a stolen laptop computer, violence or other issue, it is simply a case of working back from the event to where the suspect entered the building and then using or distributing this one image in order to identify the offender.

Over many years now the author has achieved a very high level of identification with this simple approach to CCTV that delivers the required results.

About the author
Jeff Brooks CPP is the owner of Crimewatch™ Video Pty Limited. He has installed nothing other than CCTV cameras for the past 16 years. His company owns in excess of 2000 high end cameras that are in operational use along with the associated matrix switching and recording infrastructure. In the last 3 years he has installed over 200 Terabytes of digital recording hard disk capacity.