printer friendly version

Supporting CCTV intelligence with small and big data

Intelligence functions can exist without CCTV and have done for centuries. Similarly, CCTV can function without an intelligence function, and many sites operate in a vacuum, looking only at what is in front of them. I’ve written previously about how intelligence-driven surveillance can have a major impact on CCTV system performance, and indeed intelligence has been a theme in many of my articles (https://www.securitysa.com/14414r).

The increasing development of AI, and its role in enhancing investigation-led surveillance, and the increasing capacity of CCTV control rooms and local analysts to deliver data in return can hugely increase the synergy between these two roles.

The interaction between intelligence and CCTV control room functions is fundamental to the ultimate success of both. The control room should be capturing data, including not only CCTV footage, but also information from the community, guards, response personnel, access control information, and incident details about infrastructure and modes operandi.

All this goes to create what I refer to as local or ‘small data’ that tends to be site specific. The type and quality of this data plays an important role when using it within a site (such as looking for a repeat offender), but also the way in which this is shared with ‘big data’ sources, which reconcile this information in the context of the wider society. Broader intelligence and investigations, increasingly led by AI, reconcile information from a variety of sources to provide better information on emerging threats and risk conditions. This can potentially cover the suspects, methods, times and even locations where one can look for threats. It also increases the potential profile information on the suspects, and aids in the immediate and follow-up responses to incidents.

Managing the CCTV and intelligence relationship

I wrote a number of years ago about the importance of defining and managing the relationship between CCTV operations and intelligence functions, and AI and the advent of big data has not changed that. This includes a need for:

• Defining the nature of the relationship and responsibilities of the two parties – how the parties and management can work together to maximise the benefits in a structured manner.

• What information needs to be passed on within this relationship, the quality of information required, and what may be kept confidential; the sooner this is defined, the quicker potential issues in relationships can be avoided.

• What kinds of indicators and event information are required from each party – ie., the kinds of behaviour/actions/details that CCTV need to be aware of to pass on, and the kind of things intelligence need to give surveillance access to, to make them more effective.

• A common awareness of body language and profiling, and its role in surveillance and intelligence management.

• Data coding that everybody can agree on to ensure that information generated can be incorporated into the systems of the respective parties.

• An intelligence interface that is effective and allows one to push relevant information, interrogate the system easily, and establish and illustrate patterns and relationships between different elements.

I’ve seen a number of attempts at control room data collection fail miserably due to poor information collection, inappropriate data, poor classification, and inability to transform collected data into usable information. There needs to be a focus on gathering appropriate and quality information. One of the starting points is to have a suitable awareness of your own CCTV system performance though system reporting functions.

Most modern VMS software and alarm systems have extensive report types available that summarise and highlight the weaknesses in system performance, key concerns, and success points, and incident information. All too often, management fail to pay attention to the fundamentals of the system which is impacting on other security staff and functions, and potentially limiting the ability of systems to deliver an appropriate service capacity. Often, it is only when a crisis hits that these kinds of things are addressed.

With CCTV, AI can analyse only what it sees on camera. With the volume of cameras increasing and prices coming down substantially to promote this process even more, we can expect to see many more cameras, even where they are of perhaps lesser quality or don’t have exactly the right views. AI will allow us to potentially pull as much information as possible from this range of cameras.

However, bear in mind that the PTZ camera is still the ultimate tool to gather close up and specific information about an incident, who is involved and other answers to the why, what, when, who, where and how questions. And the quality of footage that the PTZ is going to gather is going to be dependent on the cameramanship and intent of the operator.

Similarly, the operator is going to be the one to view analytics or AI notifications on camera, and make judgements on whether it is a valid concern or false alarm. The operators become the filters and one of the primary influences that will determine the relevance and quality of small and big data used subsequently for intelligence.

The information gathered by the CCTV control room needs to meet a few criteria to determine how well suited it is for integration into intelligence databases. These could be seen as follows:

• Quality of information: Is it of sufficient image quality, is it from a trusted source, is it up to date, is it free from discriminator effects, can we allocate a confidence level to it, and is it free from tampering?

• Format of information: Is it in a format that is easy to use and interpret, is it updated appropriately, is it accurate and free from false alarms, is it usable and avoiding overload?

• Verification: Verification of the accuracy of information should occur as close as possible to the source, should include confirmation through a convergence of finding from different sources or technology, and should be treated within a set of standard protocols to ensure quality adherence is maintained.

• Decision making: The data should provide a basis for decision making and acting on the decision, considerations such as weighting should be considered, and is the cost of outcomes reflected in the integrated data analysis?

Leveraging intelligence information can be key to CCTV, and much of it can start in the control room. As we move to machine vision-based approaches to understand the content of images or video, we can expect far more to come through from our control room sources. One of the key reasons is simply that the AI will be reviewing and refining information on all cameras in the background. With true AI in-depth processing of images, vector analysis, feature extraction, classification, behaviour modelling and linking, we can expect far more information to come through into the local databases, which in turn provides a better springboard into accessing the big data sources.

The better trained that operators are in observation, awareness and behavioural analysis, the higher quality we could expect the process to produce. One can then be more confident about using and integrating this small data into the broader networks and data access that makes up the big data sources. From there, we can get into the predictive modelling using real-time and historical big data to respond to live threats and extend the length of our follow up as more information flows into the control room. More generally, we can be prepared through an awareness of emerging threats, tactics, and how they impact on our vulnerabilities.

One of the fundamental challenges facing organisations in this process is the sharing of data gathered at these local levels, to allow integration that the company next to you, down the road, across the city or country, and even internationally may be able to use if it is accessed in big data environments. A reluctance to admit they have been targeted, being averse to publicity that may come through, being possessive about data and information, or acting on a silo basis, may all be factors that impact on this reluctance to share information.

This is something experienced when wanting to get similar information about company cybercrime experiences. The need to comply with legislation relating to personal information may also play a big part in the extent to which information is divulged. However, acting in concert to produce quality small data reservoirs that integrate into the big data universe in a managed way can lead to vast improvements at a local, industry, and even national level to produce intelligence to combat criminals and syndicates.

Dr Craig Donald.

Dr Craig Donald is a human factors specialist in security and CCTV. He is a director of Leaderware which provides instruments for the selection of CCTV operators, X-ray screeners and other security personnel in major operations around the world. He also runs CCTV Surveillance Skills and Body Language, and Advanced Surveillance Body Language courses for CCTV operators, supervisors and managers internationally, and consults on CCTV management. He can be contacted on +27 11 787 7811 or craig.donald@leaderware.com

Share this article:

Further reading: