Today’s AI-enabled cameras have much greater capabilities than previous generation cameras, so getting your money’s worth requires more and different preparation than previous camera projects may have required.
Q: Our security integrator proposed replacing our outdated cameras with new AI-enabled ones. Still, the labor cost only covers putting new cameras in the old camera locations and configuring IP addresses. Shouldn’t there be more to it?
A: Yes, there is more to it, and customers and their integrators can and should thoughtfully prepare to realize the value available from AI-enabled video surveillance systems.
The Common Challenge Today
Today’s challenge is matching the camera technology capabilities to the facility’s surveillance needs to achieve a worthwhile return on investment system, including physical security and facility operations benefits. Most end-users aren't deeply familiar with the modern capabilities of AI-enabled cameras and VMS software. After all, this is a rapidly advancing field. Conversely, most integrators don’t have full insight into the site’s physical security risks and facility operation needs. It’s not hard to match these up, but documentation and collaboration take significant forethought and time.
For earlier projects, many organizations have performed facility security risk assessments and planned the most useful camera locations and the fields of view they should include. However, most organizations have not done so. In some cases, security walkthroughs have been performed, and notes were taken regarding field-of-view coverage. Where such documentation exists, it can be a very helpful starting point.
Regardless of the current state of such documentation, the needed design planning is not hard to perform. The primary challenge isn’t the nature of the work involved—it’s finding the time to do it. Security design consultants specialize in that work, and leading design consultants can be found among the International Association of Professional Security Consultants (IAPSC) members.
The Important AI Difference
Timo Sachse, a long-time product analyst for camera manufacturer Axis Communications, provides a good six-minute explanation of the important aspects of camera-based AI in his recent article, “Edge AI: unlocking the power of edge computing.” Cameras' ability to understand a scene and categorize important details continues to advance. An important fact regarding AI-enabled cameras, especially those containing special AI chips for machine learning (ML) and deep learning (DL), is that the cameras you install today will keep advancing after installation.
This is a new element to consider in camera deployment design. It means documenting the rationale behind camera placement and field-of-view coverage is more important than ever. As the AI capabilities continue to advance, you’ll need to revisit how you apply the AI-based analytics advancements such as for alerting and alarming, triggering a higher level of image recording, activating automated responses based on scene activity – including notifications to security and/or facility personnel that can now include video images or clips as well as a description of the activity taking place or the type of violation that is occurring.
Important AI-Related Questions
Important questions to ask yourself regarding your application of video analytics are:
- What do I want to learn about the scene activity or the presence or motion of elements in the scene?
- With whom should I share that information?
- Based on the reliability of the analytics and the scene detail it provides, what kind of automated responses should be triggered, such as log entries, information sharing, and alerting and alarming?
Camera Selection and Placement
Should the location of any cameras be changed for camera upgrade projects, given the new video analytics capabilities? For example, AI-enabled analytics can provide very accurate counts of people and vehicle entries and exits. However, such accuracy is also dependent on the analytics capabilities and the visibility into the scene provided by the camera location and vertical angle of view.
For example, eye-level camera views can provide helpful video for identifying people entering and moving around in a building lobby entrance. However, an overhead camera is often required so that counting can be performed accurately, without some bodies obscuring the view of other bodies, as happens with eye-level cameras.
The customer is responsible for determining the desired information from each camera’s field of view. That forms the basis for which analytics to select. Once the analytics and field of view requirements are determined, camera model selection and choice of which analytics to apply will involve product knowledge. This is why collaboration is needed – either with a security design consultant or an experienced designer from the integrator’s team – to ensure that the model of camera selected supports the specific analytics needed as well as the field of view requirements and the range of lighting in the scene.
Getting Your Money’s Worth
Camera upgrade deployments are typically high-dollar investments. It is fiscally responsible to get all of the value available from that investment by doing a proper, well-documented design, even if the design collaboration requires an incremental increase in the project cost.
