Intelligent systems

CCTV is a useful and important tool for anyone seeking to improve security. Its very presence can be a powerful deterrent to potential criminals, even if there are no cameras inside the boxes. When correctly installed, the images it provides can prove invaluable in determining the perpetrators of crime. However, this is all rather reactive. If a crime is committed, it can be a long job to wade through hours of footage from multiple cameras to try and ascertain if the crime itself or any of its perpetrators were captured, and this is the case even if the system is digital.

Facial recognition technology allows systems to be proactive. Systems fitted with this technology can alert operators when faces matching those on a database of known individuals appear in the scene. For example, fitted at a shopping centre, the centre security manager can be alerted when known shoplifters enter. This then allows them to focus their security resources on these known offenders. The advantage here is that the security resources are already watching the most likely offenders, rather than looking at recorded footage after the fact, and confirming that the crime was carried out by a known shoplifter. Fundamentally, facial recognition built into a CCTV system brings with it a wealth of information. The system not only records images, but also identities – the faces of the people present are there for all to see, and for the system to match against its database. Obviously, the ultimate strength of the system lies in the information within your database: You can only recognise people who are in that database. That means that the majority of people scanned by the system will not be found within the database, and their faces will then be forgotten by the system. Only those with a match against the database will be brought to an operator's attention. Of course, not all crimes are committed by known individuals. A facial-recognition enabled CCTV system can greatly assist in the retroactive scanning of recorded information as well. Once a new offender's face can be shown to the system, not only should it be able to alert when that individual is seen again, it can also search through its archives to see if that person was seen in the past, and could perhaps be linked to other crimes or incidents, allowing a more complete pattern of behaviour to be established.

You may have heard that facial recognition doesn't work, or at least, not very well. Some of the first system implementations were certainly very ambitious, perhaps overly so, and have suffered because of this. However, lessons have been learned from these systems which can be brought to bear on new installations. Mostly, these revolve around the best way to implement and install them. There have also been many innovations around the algorithms that these systems use, one of the most exciting and promising being the use of three-dimensional facial recognition, as opposed to the currently more common format of two-dimensional systems. Obviously, the added dimension inherent within 3D systems offer better accuracy, and greater flexibility. They also get around some of the issues with 2D systems, such as reliance on constant ambient illumination. For all of these reasons, 3D systems are seen by many as offering the obvious future path for facial recognition systems and current take-up is very good.

Facial recognition is not a simple technology to install and use properly. Poor specification, installation, training and system use can completely remove the advantages. The intended use and environment should be discussed in detail with the manufacturers in many cases, to confirm that the system will be fit for purpose. Professional training of the system operatives should then lead to professional use. Existing case studies where all of this has occurred show many good applications with users benefiting from early adoption.